Dekomposition omsætning EL:CON total

import pandas as pd

df = pd.read_excel (r'C:\Users\jupe\Desktop\Praktik\210\210 - Forecast input.xlsx',sheet_name='Elcon total omsætning')
df.Periode = pd.to_datetime(df.Periode)
df2 = df
df = df.set_index('Periode')

from statsmodels.tsa.seasonal import seasonal_decompose
results = seasonal_decompose(df, model='additive')
trend_estimate    = results.trend
periodic_estimate = results.seasonal
residual = results.resid

import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import numpy as np
fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df)
ax.set(xlabel="År og måned",
       ylabel="Omsætning(100 mdkk)",
       title="Udvikling i omsætning")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df.index[0], 'Y')
datemax =  np.datetime64(df.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(trend_estimate)
ax.set(xlabel="År og måned",
       ylabel="Omsætning(10 mdkk)",
       title="Trend")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(trend_estimate.index[0], 'Y')
datemax =  np.datetime64(trend_estimate.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate)
ax.set(xlabel="År og måned",
       ylabel="Omsætning(10 mdkk)",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(periodic_estimate.index[0], 'Y')
datemax =  np.datetime64(periodic_estimate.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(residual)
ax.set(xlabel="År og måned",
       ylabel="Omsætning(10 mdkk)",
       title="Afvigelser")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(residual.index[0], 'Y')
datemax =  np.datetime64(residual.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

SARIMA for EL:CON afdelinger

df = pd.read_excel (r'C:\Users\jupe\Desktop\Praktik\210\210 - Forecast input.xlsx',sheet_name='SARIMA input',
                    skiprows=1)
df.Periode = pd.to_datetime(df.Periode)

import warnings
warnings.filterwarnings("ignore")
df.Periode = pd.to_datetime(df.Periode)
df.train = df.loc[df['Periode'].dt.year<2021]
df.test = df.loc[(df['Periode'].dt.year==2021)&(df['Periode'].dt.month<=11)]
df.train = df.train.set_index('Periode')
df.test = df.test.set_index('Periode')
df.forecast = df.loc[df['Periode'].dt.year>=2021]
df.forecast = df.forecast.set_index('Periode')

import pmdarima as pm
from sklearn.metrics import mean_squared_error

n_periods = 36
n_periods_index = n_periods+1
index_of_fc = pd.date_range(df.train.index[-1], periods = n_periods_index, freq='MS', closed = 'right')

modeldescription = pd.DataFrame()
forecast = pd.DataFrame()
rmse = pd.DataFrame()


for i in range (0,len(df.train.columns)):
    df.train.afdeling = df.train.iloc[:,[i]]
    df.train.afdeling=df.train.afdeling.dropna()
    smodel = pm.auto_arima(df.train.afdeling, start_p=1, start_q=1,
                           test='adf',
                           max_p=3, max_q=3, m=12,
                           start_P=0, seasonal=True,
                           d=None, D=1, trace=True,
                           error_action='ignore',  
                           suppress_warnings=True,
                           stepwise=True)
    
    # Forecast

    fitted, confint = smodel.predict(n_periods=n_periods,
                                     return_conf_int=True)
      
    # Save forecast and model performance measurements
    modeldescription[df.train.columns[i]] = pd.Series(smodel)
    Y_true = df.test.iloc[:,[i]]
    Y_pred = fitted[0:11]
    for j in range (1,len(Y_true)+1):
        forecast_errors = [Y_true[j-1:j]-Y_pred[j-1:j]]
    rmse[df.train.columns[i]] = pd.Series(mean_squared_error(Y_true,Y_pred,squared=False), index = ['RMSE'])
    forecast[df.train.columns[i]] = pd.Series(fitted, index=index_of_fc)

Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,1,1)[12]             : AIC=993.009, Time=0.12 sec
 ARIMA(0,2,0)(0,1,0)[12]             : AIC=1042.978, Time=0.01 sec
 ARIMA(1,2,0)(1,1,0)[12]             : AIC=1017.817, Time=0.07 sec
 ARIMA(0,2,1)(0,1,1)[12]             : AIC=995.154, Time=0.10 sec
 ARIMA(1,2,1)(0,1,0)[12]             : AIC=inf, Time=0.10 sec
 ARIMA(1,2,1)(1,1,1)[12]             : AIC=994.918, Time=0.28 sec
 ARIMA(1,2,1)(0,1,2)[12]             : AIC=994.867, Time=0.41 sec
 ARIMA(1,2,1)(1,1,0)[12]             : AIC=998.456, Time=0.18 sec
 ARIMA(1,2,1)(1,1,2)[12]             : AIC=996.549, Time=0.63 sec
 ARIMA(1,2,0)(0,1,1)[12]             : AIC=1014.385, Time=0.07 sec
 ARIMA(2,2,1)(0,1,1)[12]             : AIC=994.632, Time=0.16 sec
 ARIMA(1,2,2)(0,1,1)[12]             : AIC=992.512, Time=0.23 sec
 ARIMA(1,2,2)(0,1,0)[12]             : AIC=inf, Time=0.17 sec
 ARIMA(1,2,2)(1,1,1)[12]             : AIC=994.444, Time=0.31 sec
 ARIMA(1,2,2)(0,1,2)[12]             : AIC=994.401, Time=0.57 sec
 ARIMA(1,2,2)(1,1,0)[12]             : AIC=998.748, Time=0.25 sec
 ARIMA(1,2,2)(1,1,2)[12]             : AIC=996.376, Time=1.12 sec
 ARIMA(0,2,2)(0,1,1)[12]             : AIC=992.519, Time=0.13 sec
 ARIMA(2,2,2)(0,1,1)[12]             : AIC=993.649, Time=0.28 sec
 ARIMA(1,2,3)(0,1,1)[12]             : AIC=992.010, Time=0.23 sec
 ARIMA(1,2,3)(0,1,0)[12]             : AIC=inf, Time=0.23 sec
 ARIMA(1,2,3)(1,1,1)[12]             : AIC=993.931, Time=0.36 sec
 ARIMA(1,2,3)(0,1,2)[12]             : AIC=993.887, Time=0.54 sec
 ARIMA(1,2,3)(1,1,0)[12]             : AIC=inf, Time=0.45 sec
 ARIMA(1,2,3)(1,1,2)[12]             : AIC=995.890, Time=1.09 sec
 ARIMA(0,2,3)(0,1,1)[12]             : AIC=992.252, Time=0.19 sec
 ARIMA(2,2,3)(0,1,1)[12]             : AIC=993.978, Time=0.28 sec
 ARIMA(1,2,3)(0,1,1)[12] intercept   : AIC=996.310, Time=0.35 sec

Best model:  ARIMA(1,2,3)(0,1,1)[12]          
Total fit time: 8.930 seconds
Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,1,1)[12]             : AIC=965.822, Time=0.10 sec
 ARIMA(0,2,0)(0,1,0)[12]             : AIC=1012.913, Time=0.01 sec
 ARIMA(1,2,0)(1,1,0)[12]             : AIC=985.118, Time=0.06 sec
 ARIMA(0,2,1)(0,1,1)[12]             : AIC=968.487, Time=0.09 sec
 ARIMA(1,2,1)(0,1,0)[12]             : AIC=967.786, Time=0.08 sec
 ARIMA(1,2,1)(1,1,1)[12]             : AIC=967.771, Time=0.14 sec
 ARIMA(1,2,1)(0,1,2)[12]             : AIC=967.754, Time=0.23 sec
 ARIMA(1,2,1)(1,1,0)[12]             : AIC=968.113, Time=0.14 sec
 ARIMA(1,2,1)(1,1,2)[12]             : AIC=969.743, Time=0.33 sec
 ARIMA(1,2,0)(0,1,1)[12]             : AIC=981.362, Time=0.07 sec
 ARIMA(2,2,1)(0,1,1)[12]             : AIC=966.698, Time=0.15 sec
 ARIMA(1,2,2)(0,1,1)[12]             : AIC=967.965, Time=0.18 sec
 ARIMA(0,2,0)(0,1,1)[12]             : AIC=999.809, Time=0.02 sec
 ARIMA(0,2,2)(0,1,1)[12]             : AIC=966.369, Time=0.13 sec
 ARIMA(2,2,0)(0,1,1)[12]             : AIC=976.583, Time=0.10 sec
 ARIMA(2,2,2)(0,1,1)[12]             : AIC=970.737, Time=0.33 sec
 ARIMA(1,2,1)(0,1,1)[12] intercept   : AIC=967.992, Time=0.17 sec

Best model:  ARIMA(1,2,1)(0,1,1)[12]          
Total fit time: 2.332 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1055.200, Time=0.12 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1070.169, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1060.939, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1053.675, Time=0.07 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1065.375, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1055.662, Time=0.11 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=1055.640, Time=0.21 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1060.660, Time=0.06 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1056.143, Time=0.31 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1057.735, Time=0.03 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1056.695, Time=0.07 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1055.073, Time=0.06 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=1058.569, Time=0.11 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=1065.543, Time=0.09 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.328 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=997.821, Time=0.17 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1014.617, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1001.649, Time=0.06 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=996.092, Time=0.07 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1002.062, Time=0.03 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=996.339, Time=0.19 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=994.746, Time=0.24 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=995.426, Time=0.30 sec
 ARIMA(0,1,0)(0,1,2)[12]             : AIC=1003.152, Time=0.15 sec
 ARIMA(1,1,1)(0,1,2)[12]             : AIC=996.150, Time=0.54 sec
 ARIMA(0,1,2)(0,1,2)[12]             : AIC=995.203, Time=0.22 sec
 ARIMA(1,1,0)(0,1,2)[12]             : AIC=995.167, Time=0.19 sec
 ARIMA(1,1,2)(0,1,2)[12]             : AIC=996.726, Time=0.32 sec
 ARIMA(0,1,1)(0,1,2)[12] intercept   : AIC=1006.052, Time=0.33 sec

Best model:  ARIMA(0,1,1)(0,1,2)[12]          
Total fit time: 2.827 seconds
Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,1,1)[12]             : AIC=1036.183, Time=0.13 sec
 ARIMA(0,2,0)(0,1,0)[12]             : AIC=1074.743, Time=0.01 sec
 ARIMA(1,2,0)(1,1,0)[12]             : AIC=1050.719, Time=0.05 sec
 ARIMA(0,2,1)(0,1,1)[12]             : AIC=1037.491, Time=0.07 sec
 ARIMA(1,2,1)(0,1,0)[12]             : AIC=1042.144, Time=0.06 sec
 ARIMA(1,2,1)(1,1,1)[12]             : AIC=1037.898, Time=0.13 sec
 ARIMA(1,2,1)(0,1,2)[12]             : AIC=1037.804, Time=0.22 sec
 ARIMA(1,2,1)(1,1,0)[12]             : AIC=1038.153, Time=0.10 sec
 ARIMA(1,2,1)(1,1,2)[12]             : AIC=1039.726, Time=0.52 sec
 ARIMA(1,2,0)(0,1,1)[12]             : AIC=1047.785, Time=0.05 sec
 ARIMA(2,2,1)(0,1,1)[12]             : AIC=1036.580, Time=0.24 sec
 ARIMA(1,2,2)(0,1,1)[12]             : AIC=1032.038, Time=0.22 sec
 ARIMA(1,2,2)(0,1,0)[12]             : AIC=inf, Time=0.12 sec
 ARIMA(1,2,2)(1,1,1)[12]             : AIC=1033.608, Time=0.23 sec
 ARIMA(1,2,2)(0,1,2)[12]             : AIC=1033.439, Time=0.30 sec
 ARIMA(1,2,2)(1,1,0)[12]             : AIC=1035.445, Time=0.19 sec
 ARIMA(1,2,2)(1,1,2)[12]             : AIC=1035.279, Time=0.78 sec
 ARIMA(0,2,2)(0,1,1)[12]             : AIC=1030.997, Time=0.09 sec
 ARIMA(0,2,2)(0,1,0)[12]             : AIC=1040.361, Time=0.13 sec
 ARIMA(0,2,2)(1,1,1)[12]             : AIC=1032.630, Time=0.18 sec
 ARIMA(0,2,2)(0,1,2)[12]             : AIC=1032.485, Time=0.29 sec
 ARIMA(0,2,2)(1,1,0)[12]             : AIC=1034.275, Time=0.10 sec
 ARIMA(0,2,2)(1,1,2)[12]             : AIC=1034.333, Time=0.75 sec
 ARIMA(0,2,3)(0,1,1)[12]             : AIC=1035.159, Time=0.12 sec
 ARIMA(1,2,3)(0,1,1)[12]             : AIC=1035.757, Time=0.29 sec
 ARIMA(0,2,2)(0,1,1)[12] intercept   : AIC=1039.332, Time=0.13 sec

Best model:  ARIMA(0,2,2)(0,1,1)[12]          
Total fit time: 5.531 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=999.603, Time=0.10 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1014.372, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1004.728, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=998.074, Time=0.05 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1008.816, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=999.704, Time=0.11 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=999.387, Time=0.26 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1002.505, Time=0.07 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1001.147, Time=0.53 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1000.582, Time=0.06 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=994.747, Time=0.17 sec
 ARIMA(0,1,2)(0,1,0)[12]             : AIC=1007.345, Time=0.09 sec
 ARIMA(0,1,2)(1,1,1)[12]             : AIC=995.862, Time=0.17 sec
 ARIMA(0,1,2)(0,1,2)[12]             : AIC=995.240, Time=0.19 sec
 ARIMA(0,1,2)(1,1,0)[12]             : AIC=998.097, Time=0.08 sec
 ARIMA(0,1,2)(1,1,2)[12]             : AIC=997.125, Time=0.37 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=990.273, Time=0.21 sec
 ARIMA(1,1,2)(0,1,0)[12]             : AIC=inf, Time=0.18 sec
 ARIMA(1,1,2)(1,1,1)[12]             : AIC=991.795, Time=0.20 sec
 ARIMA(1,1,2)(0,1,2)[12]             : AIC=991.284, Time=0.31 sec
 ARIMA(1,1,2)(1,1,0)[12]             : AIC=995.274, Time=0.15 sec
 ARIMA(1,1,2)(1,1,2)[12]             : AIC=992.970, Time=0.57 sec
 ARIMA(2,1,2)(0,1,1)[12]             : AIC=990.991, Time=0.18 sec
 ARIMA(1,1,3)(0,1,1)[12]             : AIC=990.369, Time=0.29 sec
 ARIMA(0,1,3)(0,1,1)[12]             : AIC=994.413, Time=0.15 sec
 ARIMA(2,1,1)(0,1,1)[12]             : AIC=992.482, Time=0.11 sec
 ARIMA(2,1,3)(0,1,1)[12]             : AIC=990.270, Time=0.39 sec
 ARIMA(2,1,3)(0,1,0)[12]             : AIC=inf, Time=0.33 sec
 ARIMA(2,1,3)(1,1,1)[12]             : AIC=991.956, Time=0.51 sec
 ARIMA(2,1,3)(0,1,2)[12]             : AIC=991.150, Time=0.72 sec
 ARIMA(2,1,3)(1,1,0)[12]             : AIC=997.020, Time=0.38 sec
 ARIMA(2,1,3)(1,1,2)[12]             : AIC=992.627, Time=1.34 sec
 ARIMA(3,1,3)(0,1,1)[12]             : AIC=991.100, Time=0.40 sec
 ARIMA(3,1,2)(0,1,1)[12]             : AIC=992.248, Time=0.33 sec
 ARIMA(2,1,3)(0,1,1)[12] intercept   : AIC=992.817, Time=0.40 sec

Best model:  ARIMA(2,1,3)(0,1,1)[12]          
Total fit time: 9.514 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=949.166, Time=0.12 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=964.865, Time=0.00 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=954.775, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=947.201, Time=0.07 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=954.921, Time=0.03 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=949.129, Time=0.10 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=949.070, Time=0.24 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=952.074, Time=0.06 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=950.650, Time=0.32 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=953.360, Time=0.04 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=950.184, Time=0.08 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=949.871, Time=0.05 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=952.079, Time=0.15 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=950.721, Time=0.11 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.408 seconds
Performing stepwise search to minimize aic
 ARIMA(1,2,1)(0,1,1)[12]             : AIC=999.199, Time=0.11 sec
 ARIMA(0,2,0)(0,1,0)[12]             : AIC=1040.223, Time=0.01 sec
 ARIMA(1,2,0)(1,1,0)[12]             : AIC=1013.405, Time=0.07 sec
 ARIMA(0,2,1)(0,1,1)[12]             : AIC=1007.281, Time=0.09 sec
 ARIMA(1,2,1)(0,1,0)[12]             : AIC=1000.388, Time=0.05 sec
 ARIMA(1,2,1)(1,1,1)[12]             : AIC=1001.004, Time=0.19 sec
 ARIMA(1,2,1)(0,1,2)[12]             : AIC=1000.915, Time=0.25 sec
 ARIMA(1,2,1)(1,1,0)[12]             : AIC=1000.795, Time=0.10 sec
 ARIMA(1,2,1)(1,1,2)[12]             : AIC=1002.834, Time=0.40 sec
 ARIMA(1,2,0)(0,1,1)[12]             : AIC=1009.131, Time=0.05 sec
 ARIMA(2,2,1)(0,1,1)[12]             : AIC=1000.821, Time=0.13 sec
 ARIMA(1,2,2)(0,1,1)[12]             : AIC=999.318, Time=0.15 sec
 ARIMA(0,2,0)(0,1,1)[12]             : AIC=1030.052, Time=0.05 sec
 ARIMA(0,2,2)(0,1,1)[12]             : AIC=1005.325, Time=0.14 sec
 ARIMA(2,2,0)(0,1,1)[12]             : AIC=1009.954, Time=0.11 sec
 ARIMA(2,2,2)(0,1,1)[12]             : AIC=1001.288, Time=0.39 sec
 ARIMA(1,2,1)(0,1,1)[12] intercept   : AIC=1009.409, Time=0.20 sec

Best model:  ARIMA(1,2,1)(0,1,1)[12]          
Total fit time: 2.521 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1054.067, Time=0.15 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1065.586, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1059.288, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1053.168, Time=0.07 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1065.138, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1054.766, Time=0.13 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=1054.674, Time=0.17 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1057.510, Time=0.05 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1056.615, Time=0.35 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1054.753, Time=0.04 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1053.593, Time=0.07 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1055.506, Time=0.08 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=1055.531, Time=0.16 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=1056.184, Time=0.08 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.436 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1048.324, Time=0.07 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1067.533, Time=0.03 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1051.291, Time=0.08 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1052.020, Time=0.04 sec
 ARIMA(1,1,1)(0,1,0)[12]             : AIC=1057.182, Time=0.06 sec
 ARIMA(1,1,1)(1,1,1)[12]             : AIC=1049.993, Time=0.13 sec
 ARIMA(1,1,1)(0,1,2)[12]             : AIC=1049.666, Time=0.25 sec
 ARIMA(1,1,1)(1,1,0)[12]             : AIC=1052.558, Time=0.10 sec
 ARIMA(1,1,1)(1,1,2)[12]             : AIC=1052.005, Time=0.46 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1047.099, Time=0.05 sec
 ARIMA(1,1,0)(0,1,0)[12]             : AIC=1056.529, Time=0.01 sec
 ARIMA(1,1,0)(1,1,1)[12]             : AIC=1048.639, Time=0.08 sec
 ARIMA(1,1,0)(0,1,2)[12]             : AIC=1048.267, Time=0.18 sec
 ARIMA(1,1,0)(1,1,2)[12]             : AIC=1050.123, Time=0.34 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1053.347, Time=0.03 sec
 ARIMA(2,1,0)(0,1,1)[12]             : AIC=1048.937, Time=0.05 sec
 ARIMA(2,1,1)(0,1,1)[12]             : AIC=1049.403, Time=0.20 sec
 ARIMA(1,1,0)(0,1,1)[12] intercept   : AIC=1049.081, Time=0.06 sec

Best model:  ARIMA(1,1,0)(0,1,1)[12]          
Total fit time: 2.252 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1054.780, Time=0.08 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1082.378, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1052.687, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1053.002, Time=0.07 sec
 ARIMA(1,1,0)(0,1,0)[12]             : AIC=1073.408, Time=0.02 sec
 ARIMA(1,1,0)(2,1,0)[12]             : AIC=1053.938, Time=0.22 sec
 ARIMA(1,1,0)(1,1,1)[12]             : AIC=1053.849, Time=0.09 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1058.194, Time=0.07 sec
 ARIMA(1,1,0)(2,1,1)[12]             : AIC=1055.848, Time=0.30 sec
 ARIMA(0,1,0)(1,1,0)[12]             : AIC=1052.162, Time=0.04 sec
 ARIMA(0,1,0)(2,1,0)[12]             : AIC=inf, Time=0.16 sec
 ARIMA(0,1,0)(1,1,1)[12]             : AIC=1061.298, Time=0.06 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1065.090, Time=0.03 sec
 ARIMA(0,1,0)(2,1,1)[12]             : AIC=1063.235, Time=0.18 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1049.381, Time=0.06 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1066.100, Time=0.02 sec
 ARIMA(0,1,1)(2,1,0)[12]             : AIC=1050.428, Time=0.21 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1050.388, Time=0.09 sec
 ARIMA(0,1,1)(2,1,1)[12]             : AIC=1052.365, Time=0.25 sec
 ARIMA(1,1,1)(1,1,0)[12]             : AIC=1050.786, Time=0.08 sec
 ARIMA(0,1,2)(1,1,0)[12]             : AIC=1043.028, Time=0.08 sec
 ARIMA(0,1,2)(0,1,0)[12]             : AIC=1069.384, Time=0.04 sec
 ARIMA(0,1,2)(2,1,0)[12]             : AIC=1043.097, Time=0.26 sec
 ARIMA(0,1,2)(1,1,1)[12]             : AIC=1042.944, Time=0.13 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1047.566, Time=0.11 sec
 ARIMA(0,1,2)(2,1,1)[12]             : AIC=1044.878, Time=0.33 sec
 ARIMA(0,1,2)(1,1,2)[12]             : AIC=1044.862, Time=0.32 sec
 ARIMA(0,1,2)(0,1,2)[12]             : AIC=1044.666, Time=0.22 sec
 ARIMA(0,1,2)(2,1,2)[12]             : AIC=1046.860, Time=0.33 sec
 ARIMA(1,1,2)(1,1,1)[12]             : AIC=1044.877, Time=0.22 sec
 ARIMA(0,1,3)(1,1,1)[12]             : AIC=1049.026, Time=0.17 sec
 ARIMA(1,1,1)(1,1,1)[12]             : AIC=1051.763, Time=0.15 sec
 ARIMA(1,1,3)(1,1,1)[12]             : AIC=1051.004, Time=0.29 sec
 ARIMA(0,1,2)(1,1,1)[12] intercept   : AIC=1046.101, Time=0.18 sec

Best model:  ARIMA(0,1,2)(1,1,1)[12]          
Total fit time: 4.963 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=940.332, Time=0.12 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=952.079, Time=0.00 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=944.662, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=939.379, Time=0.08 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=939.449, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=941.203, Time=0.12 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=941.226, Time=0.17 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=939.590, Time=0.06 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=943.200, Time=0.41 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=946.941, Time=0.03 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=940.217, Time=0.08 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=944.163, Time=0.08 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=942.041, Time=0.15 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=942.808, Time=0.07 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.448 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1023.202, Time=0.12 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1037.787, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1030.361, Time=0.06 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1023.109, Time=0.08 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1030.864, Time=0.03 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1025.107, Time=0.11 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=1025.104, Time=0.19 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1027.640, Time=0.07 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1026.575, Time=0.31 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1025.814, Time=0.04 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1023.372, Time=0.08 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1025.602, Time=0.05 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=1025.203, Time=0.13 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=1027.179, Time=0.07 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.366 seconds
Performing stepwise search to minimize aic
 ARIMA(1,0,1)(0,1,1)[12] intercept   : AIC=1068.426, Time=0.10 sec
 ARIMA(0,0,0)(0,1,0)[12] intercept   : AIC=1062.887, Time=0.02 sec
 ARIMA(1,0,0)(1,1,0)[12] intercept   : AIC=1066.592, Time=0.06 sec
 ARIMA(0,0,1)(0,1,1)[12] intercept   : AIC=1066.431, Time=0.08 sec
 ARIMA(0,0,0)(0,1,0)[12]             : AIC=1072.871, Time=0.01 sec
 ARIMA(0,0,0)(1,1,0)[12] intercept   : AIC=1064.513, Time=0.02 sec
 ARIMA(0,0,0)(0,1,1)[12] intercept   : AIC=1064.577, Time=0.05 sec
 ARIMA(0,0,0)(1,1,1)[12] intercept   : AIC=1066.411, Time=0.22 sec
 ARIMA(1,0,0)(0,1,0)[12] intercept   : AIC=1064.674, Time=0.03 sec
 ARIMA(0,0,1)(0,1,0)[12] intercept   : AIC=1064.764, Time=0.03 sec
 ARIMA(1,0,1)(0,1,0)[12] intercept   : AIC=1066.691, Time=0.03 sec

Best model:  ARIMA(0,0,0)(0,1,0)[12] intercept
Total fit time: 0.652 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1028.548, Time=0.13 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1039.073, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1034.262, Time=0.04 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1028.315, Time=0.06 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1032.953, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1029.536, Time=0.14 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=1029.443, Time=0.16 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1031.082, Time=0.03 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1031.443, Time=0.39 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1030.574, Time=0.04 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1027.829, Time=0.09 sec
 ARIMA(0,1,2)(0,1,0)[12]             : AIC=1033.704, Time=0.04 sec
 ARIMA(0,1,2)(1,1,1)[12]             : AIC=1029.190, Time=0.16 sec
 ARIMA(0,1,2)(0,1,2)[12]             : AIC=1029.031, Time=0.22 sec
 ARIMA(0,1,2)(1,1,0)[12]             : AIC=1031.708, Time=0.07 sec
 ARIMA(0,1,2)(1,1,2)[12]             : AIC=1030.998, Time=0.38 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=1029.785, Time=0.14 sec
 ARIMA(0,1,3)(0,1,1)[12]             : AIC=1030.798, Time=0.13 sec
 ARIMA(1,1,3)(0,1,1)[12]             : AIC=1032.632, Time=0.27 sec
 ARIMA(0,1,2)(0,1,1)[12] intercept   : AIC=1035.310, Time=0.14 sec

Best model:  ARIMA(0,1,2)(0,1,1)[12]          
Total fit time: 2.686 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=975.219, Time=0.13 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=992.288, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=978.393, Time=0.06 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=973.225, Time=0.07 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=980.797, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=inf, Time=0.19 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=inf, Time=0.23 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=976.173, Time=0.05 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=inf, Time=0.41 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=982.926, Time=0.05 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=974.662, Time=0.08 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=975.911, Time=0.07 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=976.076, Time=0.24 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=983.462, Time=0.08 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12]          
Total fit time: 1.698 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=988.204, Time=0.09 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1012.578, Time=0.01 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=992.027, Time=0.06 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=986.425, Time=0.05 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=994.438, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=987.467, Time=0.09 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=987.517, Time=0.14 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=986.645, Time=0.04 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=989.467, Time=0.32 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=997.294, Time=0.03 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=987.000, Time=0.07 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=992.635, Time=0.06 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=986.923, Time=0.15 sec
 ARIMA(0,1,1)(0,1,1)[12] intercept   : AIC=986.290, Time=0.08 sec
 ARIMA(0,1,1)(0,1,0)[12] intercept   : AIC=993.259, Time=0.05 sec
 ARIMA(0,1,1)(1,1,1)[12] intercept   : AIC=987.599, Time=0.14 sec
 ARIMA(0,1,1)(0,1,2)[12] intercept   : AIC=987.294, Time=0.23 sec
 ARIMA(0,1,1)(1,1,0)[12] intercept   : AIC=987.007, Time=0.06 sec
 ARIMA(0,1,1)(1,1,2)[12] intercept   : AIC=989.292, Time=0.41 sec
 ARIMA(0,1,0)(0,1,1)[12] intercept   : AIC=999.825, Time=0.05 sec
 ARIMA(1,1,1)(0,1,1)[12] intercept   : AIC=988.261, Time=0.13 sec
 ARIMA(0,1,2)(0,1,1)[12] intercept   : AIC=994.644, Time=0.11 sec
 ARIMA(1,1,0)(0,1,1)[12] intercept   : AIC=994.647, Time=0.07 sec
 ARIMA(1,1,2)(0,1,1)[12] intercept   : AIC=988.778, Time=0.18 sec

Best model:  ARIMA(0,1,1)(0,1,1)[12] intercept
Total fit time: 2.675 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=962.599, Time=0.11 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=974.071, Time=0.02 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=962.580, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=963.024, Time=0.07 sec
 ARIMA(1,1,0)(0,1,0)[12]             : AIC=972.979, Time=0.01 sec
 ARIMA(1,1,0)(2,1,0)[12]             : AIC=964.517, Time=0.18 sec
 ARIMA(1,1,0)(1,1,1)[12]             : AIC=964.449, Time=0.13 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=964.962, Time=0.24 sec
 ARIMA(1,1,0)(2,1,1)[12]             : AIC=966.549, Time=0.50 sec
 ARIMA(0,1,0)(1,1,0)[12]             : AIC=959.343, Time=0.05 sec
 ARIMA(0,1,0)(2,1,0)[12]             : AIC=961.257, Time=0.12 sec
 ARIMA(0,1,0)(1,1,1)[12]             : AIC=963.019, Time=0.10 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=964.762, Time=0.04 sec
 ARIMA(0,1,0)(2,1,1)[12]             : AIC=964.936, Time=0.33 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=961.311, Time=0.05 sec
 ARIMA(1,1,1)(1,1,0)[12]             : AIC=960.345, Time=0.08 sec
 ARIMA(0,1,0)(1,1,0)[12] intercept   : AIC=963.624, Time=0.05 sec

Best model:  ARIMA(0,1,0)(1,1,0)[12]          
Total fit time: 2.118 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=985.064, Time=0.09 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=998.992, Time=0.02 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=991.308, Time=0.05 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=985.289, Time=0.08 sec
 ARIMA(1,1,1)(0,1,0)[12]             : AIC=inf, Time=0.06 sec
 ARIMA(1,1,1)(1,1,1)[12]             : AIC=986.633, Time=0.17 sec
 ARIMA(1,1,1)(0,1,2)[12]             : AIC=986.384, Time=0.27 sec
 ARIMA(1,1,1)(1,1,0)[12]             : AIC=989.898, Time=0.21 sec
 ARIMA(1,1,1)(1,1,2)[12]             : AIC=988.099, Time=0.61 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=987.247, Time=0.10 sec
 ARIMA(2,1,1)(0,1,1)[12]             : AIC=985.890, Time=0.31 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=986.297, Time=0.26 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=984.987, Time=0.05 sec
 ARIMA(0,1,0)(1,1,1)[12]             : AIC=986.605, Time=0.08 sec
 ARIMA(0,1,0)(0,1,2)[12]             : AIC=985.927, Time=0.12 sec
 ARIMA(0,1,0)(1,1,0)[12]             : AIC=988.088, Time=0.04 sec
 ARIMA(0,1,0)(1,1,2)[12]             : AIC=987.658, Time=0.37 sec
 ARIMA(0,1,0)(0,1,1)[12] intercept   : AIC=987.888, Time=0.06 sec

Best model:  ARIMA(0,1,0)(0,1,1)[12]          
Total fit time: 2.968 seconds
Performing stepwise search to minimize aic
 ARIMA(1,0,1)(0,1,1)[12] intercept   : AIC=1134.793, Time=0.24 sec
 ARIMA(0,0,0)(0,1,0)[12] intercept   : AIC=1138.810, Time=0.00 sec
 ARIMA(1,0,0)(1,1,0)[12] intercept   : AIC=1135.930, Time=0.08 sec
 ARIMA(0,0,1)(0,1,1)[12] intercept   : AIC=1136.548, Time=0.10 sec
 ARIMA(0,0,0)(0,1,0)[12]             : AIC=1148.552, Time=0.00 sec
 ARIMA(1,0,1)(0,1,0)[12] intercept   : AIC=1142.805, Time=0.05 sec
 ARIMA(1,0,1)(1,1,1)[12] intercept   : AIC=inf, Time=0.56 sec
 ARIMA(1,0,1)(0,1,2)[12] intercept   : AIC=inf, Time=0.87 sec
 ARIMA(1,0,1)(1,1,0)[12] intercept   : AIC=1136.937, Time=0.13 sec
 ARIMA(1,0,1)(1,1,2)[12] intercept   : AIC=1137.106, Time=0.73 sec
 ARIMA(1,0,0)(0,1,1)[12] intercept   : AIC=1136.833, Time=0.07 sec
 ARIMA(2,0,1)(0,1,1)[12] intercept   : AIC=1136.616, Time=0.23 sec
 ARIMA(1,0,2)(0,1,1)[12] intercept   : AIC=inf, Time=0.48 sec
 ARIMA(0,0,0)(0,1,1)[12] intercept   : AIC=1134.595, Time=0.05 sec
 ARIMA(0,0,0)(1,1,1)[12] intercept   : AIC=1135.548, Time=0.08 sec
 ARIMA(0,0,0)(0,1,2)[12] intercept   : AIC=1136.472, Time=0.13 sec
 ARIMA(0,0,0)(1,1,0)[12] intercept   : AIC=1133.558, Time=0.05 sec
 ARIMA(0,0,0)(2,1,0)[12] intercept   : AIC=1133.534, Time=0.15 sec
 ARIMA(0,0,0)(2,1,1)[12] intercept   : AIC=1133.577, Time=0.28 sec
 ARIMA(1,0,0)(2,1,0)[12] intercept   : AIC=1135.766, Time=0.20 sec
 ARIMA(0,0,1)(2,1,0)[12] intercept   : AIC=1135.663, Time=0.23 sec
 ARIMA(1,0,1)(2,1,0)[12] intercept   : AIC=1135.265, Time=0.48 sec
 ARIMA(0,0,0)(2,1,0)[12]             : AIC=1129.228, Time=0.10 sec
 ARIMA(0,0,0)(1,1,0)[12]             : AIC=1136.723, Time=0.03 sec
 ARIMA(0,0,0)(2,1,1)[12]             : AIC=1130.516, Time=0.22 sec
 ARIMA(0,0,0)(1,1,1)[12]             : AIC=1129.090, Time=0.08 sec
 ARIMA(0,0,0)(0,1,1)[12]             : AIC=1129.642, Time=0.07 sec
 ARIMA(0,0,0)(1,1,2)[12]             : AIC=1130.256, Time=0.26 sec
 ARIMA(0,0,0)(0,1,2)[12]             : AIC=1128.328, Time=0.14 sec
 ARIMA(1,0,0)(0,1,2)[12]             : AIC=1134.685, Time=0.20 sec
 ARIMA(0,0,1)(0,1,2)[12]             : AIC=1135.139, Time=0.18 sec
 ARIMA(1,0,1)(0,1,2)[12]             : AIC=1134.195, Time=0.50 sec

Best model:  ARIMA(0,0,0)(0,1,2)[12]          
Total fit time: 7.024 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1030.320, Time=0.09 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1054.827, Time=0.02 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1036.719, Time=0.08 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1028.604, Time=0.05 sec
 ARIMA(0,1,1)(0,1,0)[12]             : AIC=1039.707, Time=0.02 sec
 ARIMA(0,1,1)(1,1,1)[12]             : AIC=1029.784, Time=0.12 sec
 ARIMA(0,1,1)(0,1,2)[12]             : AIC=1029.941, Time=0.18 sec
 ARIMA(0,1,1)(1,1,0)[12]             : AIC=1028.715, Time=0.07 sec
 ARIMA(0,1,1)(1,1,2)[12]             : AIC=1031.635, Time=0.38 sec
 ARIMA(0,1,0)(0,1,1)[12]             : AIC=1039.000, Time=0.05 sec
 ARIMA(0,1,2)(0,1,1)[12]             : AIC=1026.061, Time=0.10 sec
 ARIMA(0,1,2)(0,1,0)[12]             : AIC=1038.478, Time=0.05 sec
 ARIMA(0,1,2)(1,1,1)[12]             : AIC=1027.640, Time=0.14 sec
 ARIMA(0,1,2)(0,1,2)[12]             : AIC=1027.738, Time=0.22 sec
 ARIMA(0,1,2)(1,1,0)[12]             : AIC=1026.824, Time=0.07 sec
 ARIMA(0,1,2)(1,1,2)[12]             : AIC=1029.279, Time=0.49 sec
 ARIMA(1,1,2)(0,1,1)[12]             : AIC=1024.848, Time=0.12 sec
 ARIMA(1,1,2)(0,1,0)[12]             : AIC=1036.799, Time=0.05 sec
 ARIMA(1,1,2)(1,1,1)[12]             : AIC=1026.641, Time=0.20 sec
 ARIMA(1,1,2)(0,1,2)[12]             : AIC=1026.674, Time=0.31 sec
 ARIMA(1,1,2)(1,1,0)[12]             : AIC=1026.317, Time=0.14 sec
 ARIMA(1,1,2)(1,1,2)[12]             : AIC=1028.378, Time=0.66 sec
 ARIMA(2,1,2)(0,1,1)[12]             : AIC=1020.557, Time=0.15 sec
 ARIMA(2,1,2)(0,1,0)[12]             : AIC=1036.990, Time=0.07 sec
 ARIMA(2,1,2)(1,1,1)[12]             : AIC=1022.556, Time=0.24 sec
 ARIMA(2,1,2)(0,1,2)[12]             : AIC=1022.556, Time=0.36 sec
 ARIMA(2,1,2)(1,1,0)[12]             : AIC=1024.383, Time=0.18 sec
 ARIMA(2,1,2)(1,1,2)[12]             : AIC=inf, Time=1.10 sec
 ARIMA(2,1,1)(0,1,1)[12]             : AIC=1019.582, Time=0.14 sec
 ARIMA(2,1,1)(0,1,0)[12]             : AIC=1038.329, Time=0.08 sec
 ARIMA(2,1,1)(1,1,1)[12]             : AIC=1021.552, Time=0.22 sec
 ARIMA(2,1,1)(0,1,2)[12]             : AIC=1021.552, Time=0.36 sec
 ARIMA(2,1,1)(1,1,0)[12]             : AIC=1023.570, Time=0.15 sec
 ARIMA(2,1,1)(1,1,2)[12]             : AIC=1023.407, Time=0.82 sec
 ARIMA(2,1,0)(0,1,1)[12]             : AIC=1020.153, Time=0.28 sec
 ARIMA(3,1,1)(0,1,1)[12]             : AIC=1021.220, Time=0.21 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1038.120, Time=0.04 sec
 ARIMA(3,1,0)(0,1,1)[12]             : AIC=1019.319, Time=0.13 sec
 ARIMA(3,1,0)(0,1,0)[12]             : AIC=1038.684, Time=0.05 sec
 ARIMA(3,1,0)(1,1,1)[12]             : AIC=1021.293, Time=0.19 sec
 ARIMA(3,1,0)(0,1,2)[12]             : AIC=1021.293, Time=0.22 sec
 ARIMA(3,1,0)(1,1,0)[12]             : AIC=1023.448, Time=0.14 sec
 ARIMA(3,1,0)(1,1,2)[12]             : AIC=inf, Time=0.72 sec
 ARIMA(3,1,0)(0,1,1)[12] intercept   : AIC=1020.833, Time=0.12 sec

Best model:  ARIMA(3,1,0)(0,1,1)[12]          
Total fit time: 9.646 seconds
Performing stepwise search to minimize aic
 ARIMA(1,1,1)(0,1,1)[12]             : AIC=1045.738, Time=0.08 sec
 ARIMA(0,1,0)(0,1,0)[12]             : AIC=1053.111, Time=0.02 sec
 ARIMA(1,1,0)(1,1,0)[12]             : AIC=1043.008, Time=0.07 sec
 ARIMA(0,1,1)(0,1,1)[12]             : AIC=1044.378, Time=0.05 sec
 ARIMA(1,1,0)(0,1,0)[12]             : AIC=1053.176, Time=0.02 sec
 ARIMA(1,1,0)(2,1,0)[12]             : AIC=1043.743, Time=0.16 sec
 ARIMA(1,1,0)(1,1,1)[12]             : AIC=1043.714, Time=0.08 sec
 ARIMA(1,1,0)(0,1,1)[12]             : AIC=1043.787, Time=0.05 sec
 ARIMA(1,1,0)(2,1,1)[12]             : AIC=1045.671, Time=0.27 sec
 ARIMA(0,1,0)(1,1,0)[12]             : AIC=1038.339, Time=0.03 sec
 ARIMA(0,1,0)(2,1,0)[12]             : AIC=1036.662, Time=0.12 sec
 ARIMA(0,1,0)(2,1,1)[12]             : AIC=1042.805, Time=0.16 sec
 ARIMA(0,1,0)(1,1,1)[12]             : AIC=1040.855, Time=0.07 sec
 ARIMA(0,1,1)(2,1,0)[12]             : AIC=1044.359, Time=0.16 sec
 ARIMA(1,1,1)(2,1,0)[12]             : AIC=1045.692, Time=0.16 sec
 ARIMA(0,1,0)(2,1,0)[12] intercept   : AIC=1043.339, Time=0.14 sec

Best model:  ARIMA(0,1,0)(2,1,0)[12]          
Total fit time: 1.639 seconds

pd.options.display.float_format = '{:.0f}'.format
modeldescription.index=['ModelDescription']
rmse.index=['RMSE']
forecast.result = modeldescription.append(rmse)
forecast.result = forecast.result.append(forecast)

forecast_data = forecast.result[1:] 
forecast_info = forecast.result[0:1] 
forecast_data = forecast_data.astype('int32')

Forecast Service Aarhus afdeling

df = pd.read_excel (r'C:\Users\jupe\Desktop\Praktik\210\210 - Forecast input.xlsx',sheet_name='210 - KPI')
Periode = pd.to_datetime(df.Periode)
df.Periode = pd.to_datetime(df.Periode)
df = df.set_index('Periode')

Training, test and validation split

df_train = df.iloc[:48]
df_val = df.iloc[48:56]
df_test = df.iloc[56:]

Check for stationarity

Omsætning

df_train_omsætning = df_train['Omsætning (t)']
from statsmodels.tsa.stattools import adfuller

result = adfuller(df_train_omsætning)
result[1]

0.4925797664100006

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning (mdkk)",
       title="Udvikling i omsætning")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_omsætning.index[0], 'Y')
datemax =  np.datetime64(df_train_omsætning.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_omsætning, model='addidative')
periodic_estimate_omsætning = results.seasonal

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning",
       title='Sæsoneffekt')

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_omsætning.index[0], 'Y')
datemax =  np.datetime64(df_train_omsætning.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_omsætning = df_train_omsætning-periodic_estimate_omsætning
result = adfuller(df_train_omsætning)
result[1]

0.4186683266739646

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning (mdkk)",
       title="Udvikling i omsætning")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_omsætning.index[0], 'Y')
datemax =  np.datetime64(df_train_omsætning.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_omsætning = df_train_omsætning.diff(periods = 1).dropna(axis = 0)
df_train_omsætning = df_train_omsætning.dropna(axis=0)

result = adfuller(df_train_omsætning)
result[1]

1.2348533463838644e-11

omsætning_train_min = min(df_train_omsætning)
omsætning_train_max = max(df_train_omsætning)
df_train_omsætning = (df_train_omsætning-omsætning_train_min)/(omsætning_train_max-omsætning_train_min)
df_train_omsætning = pd.DataFrame(df_train_omsætning, columns = ['Omsætning (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning",
       title="Udvikling i omsætning")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_omsætning.index[0], 'Y')
datemax =  np.datetime64(df_train_omsætning.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_omsætning = df['Omsætning (t)']
df_val_omsætning = df_val_omsætning.iloc[47:56]
val_periodisering = periodic_estimate_omsætning[11:20]
df_val_omsætning = df_val_omsætning - val_periodisering.values
df_val_omsætning = df_val_omsætning.diff(periods = 1)
df_val_omsætning = df_val_omsætning.dropna(axis=0)
df_val_omsætning = (df_val_omsætning-omsætning_train_min)/(omsætning_train_max-omsætning_train_min)
df_val_omsætning = pd.DataFrame(df_val_omsætning, columns = ['Omsætning (t)'])

df_test_omsætning = df['Omsætning (t)']
df_test_omsætning = df_test_omsætning.iloc[55:]
test_periodisering = periodic_estimate_omsætning[8:12]
df_test_omsætning = df_test_omsætning - test_periodisering.values
df_test_omsætning = df_test_omsætning.diff(periods = 1)
df_test_omsætning = df_test_omsætning.dropna(axis=0)
df_test_omsætning = (df_test_omsætning-omsætning_train_min)/(omsætning_train_max-omsætning_train_min)
df_test_omsætning = pd.DataFrame(df_test_omsætning, columns = ['Omsætning (t)'])

Vareforbrug

df_train_vareforbrug = df_train['Vareforbrug (t)']
fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug (mdkk)",
       title="Udvikling i vareforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_vareforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_vareforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_vareforbrug, model='addidative')
periodic_estimate_vareforbrug = results.seasonal

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_vareforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_vareforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_vareforbrug = df_train_vareforbrug-periodic_estimate_vareforbrug
result = adfuller(df_train_vareforbrug)
result[1]

0.01974672849389804

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug (mdkk)",
       title="Udvikling i vareforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_vareforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_vareforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_vareforbrug = df_train_vareforbrug.diff(periods = 1).dropna(axis = 0)
df_train_vareforbrug = df_train_vareforbrug.dropna(axis=0)
result = adfuller(df_train_vareforbrug)
result[1]

1.5101574505942425e-10

vareforbrug_train_min = min(df_train_vareforbrug)
vareforbrug_train_max = max(df_train_vareforbrug)
df_train_vareforbrug = (df_train_vareforbrug-vareforbrug_train_min)/(vareforbrug_train_max-vareforbrug_train_min)
df_train_vareforbrug = pd.DataFrame(df_train_vareforbrug, columns = ['Vareforbrug (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug",
       title="Udvikling i vareforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_vareforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_vareforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_vareforbrug = df['Vareforbrug (t)']
df_val_vareforbrug = df_val_vareforbrug.iloc[47:56]
val_periodisering = periodic_estimate_vareforbrug[11:20]
df_val_vareforbrug = df_val_vareforbrug - val_periodisering.values
df_val_vareforbrug = df_val_vareforbrug.diff(periods = 1)
df_val_vareforbrug = df_val_vareforbrug.dropna(axis=0)
df_val_vareforbrug = (df_val_vareforbrug-vareforbrug_train_min)/(vareforbrug_train_max-vareforbrug_train_min)
df_val_vareforbrug = pd.DataFrame(df_val_vareforbrug, columns = ['Vareforbrug (t)'])

df_test_vareforbrug = df['Vareforbrug (t)']
df_test_vareforbrug = df_test_vareforbrug.iloc[55:]
test_periodisering = periodic_estimate_vareforbrug[8:12]
df_test_vareforbrug = df_test_vareforbrug - test_periodisering.values
df_test_vareforbrug = df_test_vareforbrug.diff(periods = 1)
df_test_vareforbrug = df_test_vareforbrug.dropna(axis=0)
df_test_vareforbrug = (df_test_vareforbrug-vareforbrug_train_min)/(vareforbrug_train_max-vareforbrug_train_min)
df_test_vareforbrug = pd.DataFrame(df_test_vareforbrug, columns = ['Vareforbrug (t)'])

Lønforbrug

df_train_lønforbrug = df_train['Lønforbrug (t)']
result = adfuller(df_train_lønforbrug)
result[1]

0.97262213302789

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug (mdkk)",
       title="Udvikling i lønforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_lønforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_lønforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_lønforbrug, model='addidative')
periodic_estimate_lønforbrug = results.seasonal

df_train_lønforbrug = df_train_lønforbrug-periodic_estimate_lønforbrug
result = adfuller(df_train_lønforbrug)
result[1]

0.6079907889780972

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_lønforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_lønforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_lønforbrug = df_train_lønforbrug-periodic_estimate_lønforbrug
result = adfuller(df_train_lønforbrug)
result[1]

0.9364897554528006

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug (mdkk)",
       title="Udvikling i lønforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_lønforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_lønforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_lønforbrug = df_train_lønforbrug.diff(periods = 1).dropna(axis = 0)
df_train_lønforbrug = df_train_lønforbrug.dropna(axis=0)

result = adfuller(df_train_lønforbrug)
result[1]

9.544241725732661e-16

lønforbrug_train_min = min(df_train_lønforbrug)
lønforbrug_train_max = max(df_train_lønforbrug)
df_train_lønforbrug = (df_train_lønforbrug-lønforbrug_train_min)/(lønforbrug_train_max-lønforbrug_train_min)
df_train_lønforbrug = pd.DataFrame(df_train_lønforbrug, columns = ['Lønforbrug (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug",
       title="Udvikling i lønforbrug")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_lønforbrug.index[0], 'Y')
datemax =  np.datetime64(df_train_lønforbrug.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_lønforbrug = df['Lønforbrug (t)']
df_val_lønforbrug = df_val_lønforbrug.iloc[47:56]
val_periodisering = periodic_estimate_lønforbrug[11:20]
df_val_lønforbrug = df_val_lønforbrug - val_periodisering.values
df_val_lønforbrug = df_val_lønforbrug.diff(periods = 1)
df_val_lønforbrug = df_val_lønforbrug.dropna(axis=0)
df_val_lønforbrug = (df_val_lønforbrug-lønforbrug_train_min)/(lønforbrug_train_max-lønforbrug_train_min)
df_val_lønforbrug = pd.DataFrame(df_val_lønforbrug, columns = ['Lønforbrug (t)'])

df_test_lønforbrug = df['Lønforbrug (t)']
df_test_lønforbrug = df_test_lønforbrug.iloc[55:]
test_periodisering = periodic_estimate_lønforbrug[8:12]
df_test_lønforbrug = df_test_lønforbrug - test_periodisering.values
df_test_lønforbrug = df_test_lønforbrug.diff(periods = 1)
df_test_lønforbrug = df_test_lønforbrug.dropna(axis=0)
df_test_lønforbrug = (df_test_lønforbrug-lønforbrug_train_min)/(lønforbrug_train_max-lønforbrug_train_min)
df_test_lønforbrug = pd.DataFrame(df_test_lønforbrug, columns = ['Lønforbrug (t)'])

Faktureringsgrad

df_train_faktureringsgrad = df_train['Faktureringsgrad (t)']
result = adfuller(df_train_faktureringsgrad)
result[1]

0.0002342165980579539

faktureringsgrad_train_min = min(df_train_faktureringsgrad)
faktureringsgrad_train_max = max(df_train_faktureringsgrad)
df_train_faktureringsgrad = (df_train_faktureringsgrad-faktureringsgrad_train_min)/(faktureringsgrad_train_max-faktureringsgrad_train_min)
df_train_faktureringsgrad = pd.DataFrame(df_train_faktureringsgrad, columns = ['Faktureringsgrad (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_faktureringsgrad)
ax.set(xlabel="År og måned",
       ylabel="Faktureringsgrad",
       title="Udvikling i faktureringsgrad")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_faktureringsgrad.index[0], 'Y')
datemax =  np.datetime64(df_train_faktureringsgrad.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_faktureringsgrad = df_val['Faktureringsgrad (t)']
df_val_faktureringsgrad = (df_val_faktureringsgrad-faktureringsgrad_train_min)/(faktureringsgrad_train_max-faktureringsgrad_train_min)
df_val_faktureringsgrad = pd.DataFrame(df_val_faktureringsgrad, columns = ['Faktureringsgrad (t)'])

df_test_faktureringsgrad = df_test['Faktureringsgrad (t)']
df_test_faktureringsgrad = (df_test_faktureringsgrad-faktureringsgrad_train_min)/(faktureringsgrad_train_max-faktureringsgrad_train_min)
df_test_faktureringsgrad = pd.DataFrame(df_test_faktureringsgrad, columns = ['Faktureringsgrad (t)'])

Antal produktionsmedarbejdere

df_train_antalp = df_train['Antal produktionsmedarbejdere (t)']
result = adfuller(df_train_antalp)
result[1]

0.275744409221283

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Udvikling i antal produktionsmedarbejdere")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_antalp.index[0], 'Y')
datemax =  np.datetime64(df_train_antalp.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_antalp = np.log(df_train_antalp)
result = adfuller(df_train_antalp)
result[1]

0.17001858895926958

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Udvikling i antal produktionsmedarbejdere")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_antalp.index[0], 'Y')
datemax =  np.datetime64(df_train_antalp.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_antalp, model='addidative')
periodic_estimate_antalp = results.seasonal

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_antalp.index[0], 'Y')
datemax =  np.datetime64(df_train_antalp.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_antalp = df_train_antalp-periodic_estimate_antalp
result = adfuller(df_train_antalp)
result[1]

0.17407814170812058

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Udvikling i antal produktionsmedarbejdere")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_antalp.index[0], 'Y')
datemax =  np.datetime64(df_train_antalp.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_antalp = df_train_antalp.diff(periods = 1).dropna(axis = 0)
df_train_antalp = df_train_antalp.dropna(axis=0)

result = adfuller(df_train_antalp)
result[1]

0.00018877780640718744

antalp_train_min = min(df_train_antalp)
antalp_train_max = max(df_train_antalp)
df_train_antalp = (df_train_antalp-antalp_train_min)/(antalp_train_max-antalp_train_min)
df_train_antalp = pd.DataFrame(df_train_antalp, columns = ['Antal produktionsmedarbejdere (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Udvikling i antal produktionsmedarbejdere")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_antalp.index[0], 'Y')
datemax =  np.datetime64(df_train_antalp.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_antalp = df['Antal produktionsmedarbejdere (t)']
df_val_antalp = df_val_antalp.iloc[47:56]
df_val_antalp = np.log(df_val_antalp)
val_periodisering = periodic_estimate_antalp[11:20]
df_val_antalp = df_val_antalp - val_periodisering.values
df_val_antalp = df_val_antalp.diff(periods = 1)
df_val_antalp = df_val_antalp.dropna(axis=0)
df_val_antalp = (df_val_antalp-antalp_train_min)/(antalp_train_max-antalp_train_min)
df_val_antalp = pd.DataFrame(df_val_antalp, columns = ['Antal produktionsmedarbejdere (t)'])

df_test_antalp = df['Antal produktionsmedarbejdere (t)']
df_test_antalp = df_test_antalp.iloc[55:]
df_test_antalp = np.log(df_test_antalp)
val_periodisering = periodic_estimate_antalp[8:12]
df_test_antalp = df_test_antalp - val_periodisering.values
df_test_antalp = df_test_antalp.diff(periods = 1)
df_test_antalp = df_test_antalp.dropna(axis=0)
df_test_antalp = (df_test_antalp-antalp_train_min)/(antalp_train_max-antalp_train_min)
df_test_antalp = pd.DataFrame(df_test_antalp, columns = ['Antal produktionsmedarbejdere (t)'])

Arbejdsdage - ferie

df_train_arbejdsdage = df_train['Arbejdsdage - ferie (t)']
result = adfuller(df_train_arbejdsdage)
result[1]

8.418909248309158e-05

arbejdsdage_train_min = min(df_train_arbejdsdage)
arbejdsdage_train_max = max(df_train_arbejdsdage)
df_train_arbejdsdage = (df_train_arbejdsdage-arbejdsdage_train_min)/(arbejdsdage_train_max-arbejdsdage_train_min)
df_train_arbejdsdage = pd.DataFrame(df_train_arbejdsdage, columns = ['Arbejdsdage - ferie (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_arbejdsdage)
ax.set(xlabel="År og måned",
       ylabel="Arbejdsdage - ferie",
       title="Udvikling i arbejdsdage - ferie")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_arbejdsdage.index[0], 'Y')
datemax =  np.datetime64(df_train_arbejdsdage.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_arbejdsdage = df_val['Arbejdsdage - ferie (t)']
df_val_arbejdsdage = (df_val_arbejdsdage-arbejdsdage_train_min)/(arbejdsdage_train_max-arbejdsdage_train_min)
df_val_arbejdsdage = pd.DataFrame(df_val_arbejdsdage, columns = ['Arbejdsdage - ferie (t)'])

df_test_arbejdsdage = df_test['Arbejdsdage - ferie (t)']
df_test_arbejdsdage = (df_test_arbejdsdage-arbejdsdage_train_min)/(arbejdsdage_train_max-arbejdsdage_train_min)
df_test_arbejdsdage = pd.DataFrame(df_test_arbejdsdage, columns = ['Arbejdsdage - ferie (t)'])

Fakturerbar tid

df_train_fakturerbartid = df_train['Fakturerbar tid (t)']
result = adfuller(df_train_fakturerbartid)
result[1]

0.9242046478151071

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Udvikling i fakturerbar tid")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_fakturerbartid.index[0], 'Y')
datemax =  np.datetime64(df_train_fakturerbartid.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_fakturerbartid = np.log(df_train_fakturerbartid)
result = adfuller(df_train_fakturerbartid)
result[1]

0.837361022546167

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Udvikling i fakturerbar tid")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_fakturerbartid.index[0], 'Y')
datemax =  np.datetime64(df_train_fakturerbartid.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_fakturerbartid, model='addidative')
periodic_estimate_fakturerbartid = results.seasonal

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_fakturerbartid.index[0], 'Y')
datemax =  np.datetime64(df_train_fakturerbartid.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_fakturerbartid = df_train_fakturerbartid-periodic_estimate_fakturerbartid
result = adfuller(df_train_fakturerbartid)
result[1]

0.7489928908044364

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Udvikling i fakturerbar tid")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_fakturerbartid.index[0], 'Y')
datemax =  np.datetime64(df_train_fakturerbartid.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_fakturerbartid = df_train_fakturerbartid.diff(periods = 1)
df_train_fakturerbartid = df_train_fakturerbartid.dropna(axis=0)

result = adfuller(df_train_fakturerbartid)
result[1]

3.98147103312809e-13

fakturerbartid_train_min = min(df_train_fakturerbartid)
fakturerbartid_train_max = max(df_train_fakturerbartid)
df_train_fakturerbartid = (df_train_fakturerbartid-fakturerbartid_train_min)/(fakturerbartid_train_max-fakturerbartid_train_min)
df_train_fakturerbartid = pd.DataFrame(df_train_fakturerbartid, columns = ['Fakturerbar tid (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Udvikling i Fakturerbar tid")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_fakturerbartid.index[0], 'Y')
datemax =  np.datetime64(df_train_fakturerbartid.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_fakturerbartid = df['Fakturerbar tid (t)']
df_val_fakturerbartid = df_val_fakturerbartid.iloc[47:56]
df_val_fakturerbartid = np.log(df_val_fakturerbartid)
val_periodisering = periodic_estimate_fakturerbartid[11:20]
df_val_fakturerbartid = df_val_fakturerbartid - val_periodisering.values
df_val_fakturerbartid = df_val_fakturerbartid.diff(periods = 1)
df_val_fakturerbartid = df_val_fakturerbartid.dropna(axis=0)
df_val_fakturerbartid = (df_val_fakturerbartid-fakturerbartid_train_min)/(fakturerbartid_train_max-fakturerbartid_train_min)
df_val_fakturerbartid = pd.DataFrame(df_val_fakturerbartid, columns = ['Fakturerbar tid (t)'])

df_test_fakturerbartid = df['Fakturerbar tid (t)']
df_test_fakturerbartid = df_test_fakturerbartid.iloc[55:]
df_test_fakturerbartid = np.log(df_test_fakturerbartid)
test_periodisering = periodic_estimate_fakturerbartid[8:12]
df_test_fakturerbartid = df_test_fakturerbartid - test_periodisering.values
df_test_fakturerbartid = df_test_fakturerbartid.diff(periods = 1)
df_test_fakturerbartid = df_test_fakturerbartid.dropna(axis=0)
df_test_fakturerbartid = (df_test_fakturerbartid-fakturerbartid_train_min)/(fakturerbartid_train_max-fakturerbartid_train_min)
df_test_fakturerbartid = pd.DataFrame(df_test_fakturerbartid, columns = ['Fakturerbar tid (t)'])

Jobtimer

df_train_jobtimer = df_train['Jobtimer (t)']
result = adfuller(df_train_jobtimer)
result[1]

0.9005892686087571

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Udvikling i jobtimer")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_jobtimer.index[0], 'Y')
datemax =  np.datetime64(df_train_jobtimer.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

results = seasonal_decompose(df_train_jobtimer, model='addidative')
periodic_estimate_jobtimer = results.seasonal

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(periodic_estimate_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Sæson effekt")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_jobtimer.index[0], 'Y')
datemax =  np.datetime64(df_train_jobtimer.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_jobtimer = df_train_jobtimer-periodic_estimate_jobtimer
result = adfuller(df_train_jobtimer)
result[1]

0.9640269789920451

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Udvikling i jobtimer")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_jobtimer.index[0], 'Y')
datemax =  np.datetime64(df_train_jobtimer.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_train_jobtimer = df_train_jobtimer.diff(periods = 1).dropna(axis = 0)
df_train_jobtimer = df_train_jobtimer.dropna(axis=0)

result = adfuller(df_train_jobtimer)
result[1]

0.32967546705757433

jobtimer_train_min = min(df_train_jobtimer)
jobtimer_train_max = max(df_train_jobtimer)
df_train_jobtimer = (df_train_jobtimer-jobtimer_train_min)/(jobtimer_train_max-jobtimer_train_min)
df_train_jobtimer = pd.DataFrame(df_train_jobtimer, columns = ['Jobtimer (t)'])

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_train_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Udvikling i Jobtimer")

fmt_quarter_year = mdates.MonthLocator(interval=3)
ax.xaxis.set_major_locator(fmt_quarter_year)

fmt_month = mdates.MonthLocator()
ax.xaxis.set_minor_locator(fmt_month)

ax.xaxis.set_major_formatter(mdates.DateFormatter('%Y-%m'))

datemin = np.datetime64(df_train_jobtimer.index[0], 'Y')
datemax =  np.datetime64(df_train_jobtimer.index[-1],'Y')  + np.timedelta64(1, 'Y')
ax.set_xlim(datemin, datemax)

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

fig.autofmt_xdate()
plt.show()

df_val_jobtimer = df['Jobtimer (t)']
df_val_jobtimer = df_val_jobtimer.iloc[47:56]
val_periodisering = periodic_estimate_jobtimer[11:20]
df_val_jobtimer = df_val_jobtimer - val_periodisering.values
df_val_jobtimer = df_val_jobtimer.diff(periods = 1)
df_val_jobtimer = df_val_jobtimer.dropna(axis=0)
df_val_jobtimer = (df_val_jobtimer-jobtimer_train_min)/(jobtimer_train_max-jobtimer_train_min)
df_val_jobtimer = pd.DataFrame(df_val_jobtimer, columns = ['Jobtimer (t)'])

df_test_jobtimer = df['Jobtimer (t)']
df_test_jobtimer = df_test_jobtimer.iloc[55:]
test_periodisering = periodic_estimate_jobtimer[8:12]
df_test_jobtimer = df_test_jobtimer - test_periodisering.values
df_test_jobtimer = df_test_jobtimer.diff(periods = 1)
df_test_jobtimer = df_test_jobtimer.dropna(axis=0)
df_test_jobtimer = (df_test_jobtimer-jobtimer_train_min)/(jobtimer_train_max-jobtimer_train_min)
df_test_jobtimer = pd.DataFrame(df_test_jobtimer, columns = ['Jobtimer (t)'])

Samle træning, validerings og test datasæt

df_train_copy = df_train
df_train = df_train_omsætning.merge(df_train_vareforbrug, left_index =True, right_index = True)
df_train = df_train.merge(df_train_lønforbrug, left_index =True, right_index = True)
df_train = df_train.merge(df_train_faktureringsgrad, left_index =True, right_index = True)
df_train = df_train.merge(df_train_antalp, left_index =True, right_index = True)
df_train = df_train.merge(df_train_arbejdsdage, left_index =True, right_index = True)
df_train = df_train.merge(df_train_fakturerbartid, left_index =True, right_index = True)
df_train = df_train.merge(df_train_jobtimer, left_index =True, right_index = True)

df_val_copy = df_val
df_val = df_val_omsætning.merge(df_val_vareforbrug, left_index =True, right_index = True)
df_val = df_val.merge(df_val_lønforbrug, left_index =True, right_index = True)
df_val = df_val.merge(df_val_faktureringsgrad, left_index =True, right_index = True)
df_val = df_val.merge(df_val_antalp, left_index =True, right_index = True)
df_val = df_val.merge(df_val_arbejdsdage, left_index =True, right_index = True)
df_val = df_val.merge(df_val_fakturerbartid, left_index =True, right_index = True)
df_val = df_val.merge(df_val_jobtimer, left_index =True, right_index = True)

df_test_copy = df_test
df_test = df_test_omsætning.merge(df_test_vareforbrug, left_index =True, right_index = True)
df_test = df_test.merge(df_test_lønforbrug, left_index =True, right_index = True)
df_test = df_test.merge(df_test_faktureringsgrad, left_index =True, right_index = True)
df_test = df_test.merge(df_test_antalp, left_index =True, right_index = True)
df_test = df_test.merge(df_test_arbejdsdage, left_index =True, right_index = True)
df_test = df_test.merge(df_test_fakturerbartid, left_index =True, right_index = True)
df_test = df_test.merge(df_test_jobtimer, left_index =True, right_index = True)

Valg af KPI

df_train_kpi = df_train

df_train_kpi['Omsætning (t-1)'] = df_train_kpi['Omsætning (t)'].shift(1)
df_train_kpi['Vareforbrug (t-1)'] = df_train_kpi['Vareforbrug (t)'].shift(1)
df_train_kpi['Lønforbrug (t-1)'] = df_train_kpi['Lønforbrug (t)'].shift(1)
df_train_kpi['Faktureringsgrad (t-1)'] = df_train_kpi['Faktureringsgrad (t)'].shift(1)
df_train_kpi['Antal produktionsmedarbejdere (t-1)'] = df_train_kpi['Antal produktionsmedarbejdere (t)'].shift(1)
df_train_kpi['Arbejdsdage - ferie (t-1)'] = df_train_kpi['Arbejdsdage - ferie (t)'].shift(1)
df_train_kpi['Fakturerbar tid (t-1)'] = df_train_kpi['Fakturerbar tid (t)'].shift(1)
df_train_kpi['Jobtimer (t-1)'] = df_train_kpi['Jobtimer (t)'].shift(1)
df_train_kpi = df_train_kpi.dropna(axis=0)

pd.options.mode.chained_assignment = None  
df_train_kpi['År (t)'] = df_train_kpi.index.year
df_train_kpi['Måned (t)'] = df_train_kpi.index.month
dummies = pd.get_dummies(df_train_kpi['Måned (t)'])
df_train_kpi = df_train_kpi.drop('Måned (t)',axis = 1)
dummies = dummies.drop(1,axis = 1)
dummies.columns = ['Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)','August (t)','September (t)',
                   'Oktober (t)','November (t)','December (t)']
df_train_kpi = df_train_kpi.join(dummies)

df_monthly_dummies = pd.DataFrame(df.index)
df_monthly_dummies = df_monthly_dummies.set_index('Periode')
df_monthly_dummies['Måned (t)'] = df_monthly_dummies.index.month

dummies = pd.get_dummies(df_monthly_dummies['Måned (t)'])
df_monthly_dummies = df_monthly_dummies.drop('Måned (t)',axis = 1)
dummies = dummies.drop(1,axis = 1)
dummies.columns = ['Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)','August (t)','September (t)',
                   'Oktober (t)','November (t)','December (t)']
df_monthly_dummies = df_monthly_dummies.join(dummies)

Omsætning

df_train_kpi_omsætning = df_train_kpi.drop(['Arbejdsdage - ferie (t-1)','Faktureringsgrad (t-1)',
                                            'Antal produktionsmedarbejdere (t-1)'],axis = 1)
df_train_kpi_omsætning.corr()['Omsætning (t)'][:].sort_values()

Vareforbrug (t-1)                   -0
Omsætning (t-1)                     -0
Fakturerbar tid (t-1)               -0
Jobtimer (t-1)                      -0
August (t)                          -0
Lønforbrug (t-1)                    -0
Maj (t)                             -0
April (t)                           -0
December (t)                        -0
November (t)                        -0
Juni (t)                            -0
Februar (t)                          0
Faktureringsgrad (t)                 0
Oktober (t)                          0
År (t)                               0
September (t)                        0
Juli (t)                             0
Marts (t)                            0
Antal produktionsmedarbejdere (t)    0
Arbejdsdage - ferie (t)              0
Lønforbrug (t)                       0
Fakturerbar tid (t)                  0
Jobtimer (t)                         0
Vareforbrug (t)                      1
Omsætning (t)                        1
Name: Omsætning (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Omsætning = df_train_kpi_omsætning['Omsætning (t)']
X_Omsætning = df_train_kpi_omsætning.drop('Omsætning (t)',axis = 1)
X_Omsætning = X_Omsætning.drop(['Fakturerbar tid (t)','Fakturerbar tid (t-1)'],axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Omsætning, Y_Omsætning)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Omsætning.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Vareforbrug (t)
Vareforbrug (t-1)
Omsætning (t-1)
Jobtimer (t)

df_train_omsætning_variables = df_train_kpi[['Omsætning (t)','Vareforbrug (t)','Vareforbrug (t-1)',
                                             'Omsætning (t-1)','Jobtimer (t)']]

df_val_omsætning_variables1 = pd.DataFrame(df_train[['Omsætning (t)','Vareforbrug (t)','Jobtimer (t)']])
df_val_omsætning_variables1 = pd.DataFrame(df_val_omsætning_variables1.iloc[-1]).transpose()
df_val_omsætning_variables2 = pd.DataFrame(df_val[['Omsætning (t)','Vareforbrug (t)','Jobtimer (t)']])
df_val_omsætning_variables = df_val_omsætning_variables1.append(df_val_omsætning_variables2)
df_val_omsætning_variables['Omsætning (t-1)'] = df_val_omsætning_variables['Omsætning (t)'].shift(1)
df_val_omsætning_variables['Vareforbrug (t-1)'] = df_val_omsætning_variables['Vareforbrug (t)'].shift(1)
df_val_omsætning_variables = df_val_omsætning_variables.dropna(axis = 0)

df_test_omsætning_variables1 = pd.DataFrame(df_val[['Omsætning (t)','Vareforbrug (t)','Jobtimer (t)']])
df_test_omsætning_variables1 = pd.DataFrame(df_test_omsætning_variables1.iloc[-1]).transpose()
df_test_omsætning_variables2 = pd.DataFrame(df_test[['Omsætning (t)','Vareforbrug (t)','Jobtimer (t)']])
df_test_omsætning_variables = df_test_omsætning_variables1.append(df_test_omsætning_variables2)
df_test_omsætning_variables['Omsætning (t-1)'] = df_test_omsætning_variables['Omsætning (t)'].shift(1)
df_test_omsætning_variables['Vareforbrug (t-1)'] = df_test_omsætning_variables['Vareforbrug (t)'].shift(1)
df_test_omsætning_variables = df_test_omsætning_variables.dropna(axis = 0)

Vareforbrug

df_train_kpi_vareforbrug = df_train_kpi.drop(['Omsætning (t)','Arbejdsdage - ferie (t-1)','Faktureringsgrad (t-1)',
                                              'Antal produktionsmedarbejdere (t-1)'],axis = 1)
df_train_kpi_vareforbrug.corr()['Vareforbrug (t)'][:].sort_values()

Vareforbrug (t-1)                   -0
Omsætning (t-1)                     -0
Jobtimer (t-1)                      -0
August (t)                          -0
Fakturerbar tid (t-1)               -0
December (t)                        -0
Lønforbrug (t-1)                    -0
April (t)                           -0
Maj (t)                             -0
Oktober (t)                         -0
Juni (t)                            -0
Faktureringsgrad (t)                -0
Februar (t)                          0
År (t)                               0
November (t)                         0
September (t)                        0
Arbejdsdage - ferie (t)              0
Marts (t)                            0
Antal produktionsmedarbejdere (t)    0
Juli (t)                             0
Fakturerbar tid (t)                  0
Lønforbrug (t)                       0
Jobtimer (t)                         0
Vareforbrug (t)                      1
Name: Vareforbrug (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Vareforbrug = df_train_kpi_vareforbrug['Vareforbrug (t)']
X_Vareforbrug = df_train_kpi_vareforbrug.drop('Vareforbrug (t)',axis = 1)
X_Vareforbrug = X_Vareforbrug.drop(['Fakturerbar tid (t)','Fakturerbar tid (t-1)'],axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Vareforbrug, Y_Vareforbrug)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Vareforbrug.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Vareforbrug (t-1)
Omsætning (t-1)
Antal produktionsmedarbejdere (t)
Jobtimer (t)
Lønforbrug (t)

df_train_vareforbrug_variables = df_train_kpi[['Vareforbrug (t)','Vareforbrug (t-1)','Omsætning (t-1)',
                                               'Antal produktionsmedarbejdere (t)','Jobtimer (t)','Lønforbrug (t)']]

df_val_vareforbrug_variables1 = pd.DataFrame(df_train[['Vareforbrug (t)','Omsætning (t)','Antal produktionsmedarbejdere (t)',
                                                      'Jobtimer (t)','Lønforbrug (t)']])
df_val_vareforbrug_variables1 = pd.DataFrame(df_val_vareforbrug_variables1.iloc[-1]).transpose()
df_val_vareforbrug_variables2 = pd.DataFrame(df_val[['Vareforbrug (t)','Omsætning (t)','Antal produktionsmedarbejdere (t)',
                                                      'Jobtimer (t)','Lønforbrug (t)']])
df_val_vareforbrug_variables = df_val_vareforbrug_variables1.append(df_val_vareforbrug_variables2)
df_val_vareforbrug_variables['Vareforbrug (t-1)'] = df_val_vareforbrug_variables['Vareforbrug (t)'].shift(1)
df_val_vareforbrug_variables['Omsætning (t-1)'] = df_val_vareforbrug_variables['Omsætning (t)'].shift(1)
df_val_vareforbrug_variables = df_val_vareforbrug_variables.dropna(axis = 0)

df_test_vareforbrug_variables1 = pd.DataFrame(df_val[['Vareforbrug (t)','Omsætning (t)','Antal produktionsmedarbejdere (t)',
                                                      'Jobtimer (t)','Lønforbrug (t)']])
df_test_vareforbrug_variables1 = pd.DataFrame(df_test_vareforbrug_variables1.iloc[-1]).transpose()
df_test_vareforbrug_variables2 = pd.DataFrame(df_test[['Vareforbrug (t)','Omsætning (t)','Antal produktionsmedarbejdere (t)',
                                                      'Jobtimer (t)','Lønforbrug (t)']])
df_test_vareforbrug_variables = df_test_vareforbrug_variables1.append(df_test_vareforbrug_variables2)
df_test_vareforbrug_variables['Vareforbrug (t-1)'] = df_test_vareforbrug_variables['Vareforbrug (t)'].shift(1)
df_test_vareforbrug_variables['Omsætning (t-1)'] = df_test_vareforbrug_variables['Omsætning (t)'].shift(1)
df_test_vareforbrug_variables = df_test_vareforbrug_variables.dropna(axis = 0)

Lønforbrug

df_train_kpi_lønforbrug = df_train_kpi.drop(['Vareforbrug (t)','Omsætning (t)','Arbejdsdage - ferie (t-1)',
                                             'Faktureringsgrad (t-1)','Antal produktionsmedarbejdere (t-1)'],axis = 1)
df_train_kpi_lønforbrug.corr()['Lønforbrug (t)'][:].sort_values()

Lønforbrug (t-1)                    -0
Vareforbrug (t-1)                   -0
Jobtimer (t-1)                      -0
September (t)                       -0
Omsætning (t-1)                     -0
Oktober (t)                         -0
August (t)                          -0
Fakturerbar tid (t-1)               -0
Faktureringsgrad (t)                -0
Maj (t)                             -0
Arbejdsdage - ferie (t)              0
April (t)                            0
År (t)                               0
Juli (t)                             0
Antal produktionsmedarbejdere (t)    0
Marts (t)                            0
Juni (t)                             0
November (t)                         0
Februar (t)                          0
December (t)                         0
Fakturerbar tid (t)                  0
Jobtimer (t)                         1
Lønforbrug (t)                       1
Name: Lønforbrug (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Lønforbrug = df_train_kpi_lønforbrug['Lønforbrug (t)']
X_Lønforbrug = df_train_kpi_lønforbrug.drop('Lønforbrug (t)',axis = 1)
X_Lønforbrug = X_Lønforbrug.drop(['Fakturerbar tid (t)','Fakturerbar tid (t-1)'],axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Lønforbrug, Y_Lønforbrug)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Lønforbrug.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Jobtimer (t)
Lønforbrug (t-1)
Arbejdsdage - ferie (t)
Antal produktionsmedarbejdere (t)

df_train_lønforbrug_variables = df_train_kpi[['Lønforbrug (t)','Jobtimer (t)','Lønforbrug (t-1)','Arbejdsdage - ferie (t)',
                                              'Antal produktionsmedarbejdere (t)']]

df_val_lønforbrug_variables1 = pd.DataFrame(df_train[['Lønforbrug (t)','Jobtimer (t)','Arbejdsdage - ferie (t)',
                                              'Antal produktionsmedarbejdere (t)']])
df_val_lønforbrug_variables1 = pd.DataFrame(df_val_lønforbrug_variables1.iloc[-1]).transpose()
df_val_lønforbrug_variables2 = pd.DataFrame(df_val[['Lønforbrug (t)','Jobtimer (t)','Arbejdsdage - ferie (t)',
                                              'Antal produktionsmedarbejdere (t)']])
df_val_lønforbrug_variables = df_val_lønforbrug_variables1.append(df_val_lønforbrug_variables2)
df_val_lønforbrug_variables['Lønforbrug (t-1)'] = df_val_lønforbrug_variables['Lønforbrug (t)'].shift(1)
df_val_lønforbrug_variables = df_val_lønforbrug_variables.dropna(axis = 0)

df_test_lønforbrug_variables1 = pd.DataFrame(df_val[['Lønforbrug (t)','Jobtimer (t)','Arbejdsdage - ferie (t)',
                                              'Antal produktionsmedarbejdere (t)']])
df_test_lønforbrug_variables1 = pd.DataFrame(df_test_lønforbrug_variables1.iloc[-1]).transpose()
df_test_lønforbrug_variables2 = pd.DataFrame(df_test[['Lønforbrug (t)','Jobtimer (t)','Arbejdsdage - ferie (t)',
                                              'Antal produktionsmedarbejdere (t)']])
df_test_lønforbrug_variables = df_test_lønforbrug_variables1.append(df_test_lønforbrug_variables2)
df_test_lønforbrug_variables['Lønforbrug (t-1)'] = df_test_lønforbrug_variables['Lønforbrug (t)'].shift(1)
df_test_lønforbrug_variables = df_test_lønforbrug_variables.dropna(axis = 0)

Faktureringsgrad

df_train_kpi_faktureringsgrad = df_train_kpi.drop(['Vareforbrug (t)','Omsætning (t)','Arbejdsdage - ferie (t-1)',
                                                   'Lønforbrug (t)','Antal produktionsmedarbejdere (t-1)',
                                                   'Jobtimer (t)'],axis = 1)
df_train_kpi_faktureringsgrad.corr()['Faktureringsgrad (t)'][:].sort_values()

Februar (t)                         -0
Arbejdsdage - ferie (t)             -0
September (t)                       -0
December (t)                        -0
Oktober (t)                         -0
Marts (t)                           -0
Fakturerbar tid (t)                 -0
November (t)                         0
Antal produktionsmedarbejdere (t)    0
Maj (t)                              0
Jobtimer (t-1)                       0
August (t)                           0
Fakturerbar tid (t-1)                0
Omsætning (t-1)                      0
Juni (t)                             0
Vareforbrug (t-1)                    0
År (t)                               0
Juli (t)                             0
Lønforbrug (t-1)                     0
April (t)                            0
Faktureringsgrad (t-1)               0
Faktureringsgrad (t)                 1
Name: Faktureringsgrad (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Faktureringsgrad = df_train_kpi_faktureringsgrad['Faktureringsgrad (t)']
X_Faktureringsgrad = df_train_kpi_faktureringsgrad.drop('Faktureringsgrad (t)',axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Faktureringsgrad, Y_Faktureringsgrad)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Faktureringsgrad.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Faktureringsgrad (t-1)
Måned (t)
Arbejdsdage - ferie (t)

df_train_faktureringsgrad_variables = df_train_kpi[['Faktureringsgrad (t)','Faktureringsgrad (t-1)','Arbejdsdage - ferie (t)',
                                                    'Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)',
                                                    'August (t)','September (t)','Oktober (t)','November (t)','December (t)']]

df_val_faktureringsgrad_variables1 = pd.DataFrame(df_train[['Faktureringsgrad (t)','Arbejdsdage - ferie (t)']])
df_val_faktureringsgrad_variables1 = pd.DataFrame(df_val_faktureringsgrad_variables1.iloc[-1]).transpose()
df_val_faktureringsgrad_variables2 = pd.DataFrame(df_val[['Faktureringsgrad (t)','Arbejdsdage - ferie (t)']])
df_val_faktureringsgrad_variables = df_val_faktureringsgrad_variables1.append(df_val_faktureringsgrad_variables2)
df_val_faktureringsgrad_variables['Faktureringsgrad (t-1)'] = df_val_faktureringsgrad_variables['Faktureringsgrad (t)'].shift(1)
df_val_faktureringsgrad_variables = df_val_faktureringsgrad_variables.dropna(axis = 0)
df_val_faktureringsgrad_variables = df_val_faktureringsgrad_variables.merge(df_monthly_dummies, left_index = True,
                                                                            right_index = True)

df_test_faktureringsgrad_variables1 = pd.DataFrame(df_val[['Faktureringsgrad (t)','Arbejdsdage - ferie (t)']])
df_test_faktureringsgrad_variables1 = pd.DataFrame(df_test_faktureringsgrad_variables1.iloc[-1]).transpose()
df_test_faktureringsgrad_variables2 = pd.DataFrame(df_test[['Faktureringsgrad (t)','Arbejdsdage - ferie (t)']])
df_test_faktureringsgrad_variables = df_test_faktureringsgrad_variables1.append(df_test_faktureringsgrad_variables2)
df_test_faktureringsgrad_variables['Faktureringsgrad (t-1)'] = df_test_faktureringsgrad_variables['Faktureringsgrad (t)'].shift(1)
df_test_faktureringsgrad_variables = df_test_faktureringsgrad_variables.dropna(axis = 0)
df_test_faktureringsgrad_variables = df_test_faktureringsgrad_variables.merge(df_monthly_dummies, left_index = True,
                                                                            right_index = True)

Fakturerbar tid

df_train_kpi_fakturerbartid= df_train_kpi.drop(['Vareforbrug (t)','Omsætning (t)','Arbejdsdage - ferie (t-1)',
                                                'Lønforbrug (t)','Antal produktionsmedarbejdere (t-1)',
                                                'Jobtimer (t)','Faktureringsgrad (t)'],axis = 1)
df_train_kpi_fakturerbartid.corr()['Fakturerbar tid (t)'][:].sort_values()

Jobtimer (t-1)                      -0
Fakturerbar tid (t-1)               -0
April (t)                           -0
Omsætning (t-1)                     -0
Vareforbrug (t-1)                   -0
August (t)                          -0
Lønforbrug (t-1)                    -0
November (t)                        -0
Oktober (t)                         -0
Juni (t)                            -0
Februar (t)                         -0
Juli (t)                             0
Faktureringsgrad (t-1)               0
Maj (t)                              0
September (t)                        0
December (t)                         0
År (t)                               0
Marts (t)                            0
Antal produktionsmedarbejdere (t)    0
Arbejdsdage - ferie (t)              0
Fakturerbar tid (t)                  1
Name: Fakturerbar tid (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Fakturerbartid = df_train_kpi_fakturerbartid['Fakturerbar tid (t)']
X_Fakturerbartid = df_train_kpi_fakturerbartid.drop('Fakturerbar tid (t)',axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Fakturerbartid, Y_Fakturerbartid)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Fakturerbartid.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Antal produktionsmedarbejdere (t)
Arbejdsdage - ferie (t)
Fakturerbar tid (t-1)

df_train_fakturerbartid_variables = df_train_kpi[['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)',
                                                  'Arbejdsdage - ferie (t)','Fakturerbar tid (t-1)']]

df_val_fakturerbartid_variables1 = pd.DataFrame(df_train[['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)',
                                                          'Arbejdsdage - ferie (t)']])
df_val_fakturerbartid_variables1 = pd.DataFrame(df_val_fakturerbartid_variables1.iloc[-1]).transpose()
df_val_fakturerbartid_variables2 = pd.DataFrame(df_val[['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)',
                                                          'Arbejdsdage - ferie (t)']])
df_val_fakturerbartid_variables = df_val_fakturerbartid_variables1.append(df_val_fakturerbartid_variables2)
df_val_fakturerbartid_variables['Fakturerbar tid (t-1)'] = df_val_fakturerbartid_variables['Fakturerbar tid (t)'].shift(1)
df_val_fakturerbartid_variables = df_val_fakturerbartid_variables.dropna(axis = 0)

df_test_fakturerbartid_variables1 = pd.DataFrame(df_val[['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)',
                                                          'Arbejdsdage - ferie (t)']])
df_test_fakturerbartid_variables1 = pd.DataFrame(df_test_fakturerbartid_variables1.iloc[-1]).transpose()
df_test_fakturerbartid_variables2 = pd.DataFrame(df_test[['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)',
                                                          'Arbejdsdage - ferie (t)']])
df_test_fakturerbartid_variables = df_test_fakturerbartid_variables1.append(df_test_fakturerbartid_variables2)
df_test_fakturerbartid_variables['Fakturerbar tid (t-1)'] = df_test_fakturerbartid_variables['Fakturerbar tid (t)'].shift(1)
df_test_fakturerbartid_variables = df_test_fakturerbartid_variables.dropna(axis = 0)

Antal produktionsmedarbejdere

df_train_kpi_antalp= df_train_kpi.drop(['Vareforbrug (t)','Omsætning (t)','Arbejdsdage - ferie (t-1)','Lønforbrug (t)',
                                        'Fakturerbar tid (t)','Jobtimer (t)','Faktureringsgrad (t)'],axis = 1)
df_train_kpi_antalp.corr()['Antal produktionsmedarbejdere (t)'][:].sort_values()

August (t)                            -0
Arbejdsdage - ferie (t)               -0
Juni (t)                              -0
September (t)                         -0
Februar (t)                           -0
Oktober (t)                           -0
Marts (t)                             -0
November (t)                           0
Maj (t)                                0
December (t)                           0
April (t)                              0
Vareforbrug (t-1)                      0
Faktureringsgrad (t-1)                 0
Lønforbrug (t-1)                       0
Omsætning (t-1)                        0
Fakturerbar tid (t-1)                  0
År (t)                                 0
Jobtimer (t-1)                         0
Juli (t)                               0
Antal produktionsmedarbejdere (t-1)    0
Antal produktionsmedarbejdere (t)      1
Name: Antal produktionsmedarbejdere (t), dtype: float64

from sklearn.ensemble import RandomForestRegressor
from sklearn.inspection import permutation_importance
Y_Antalp = df_train_kpi_antalp['Antal produktionsmedarbejdere (t)']
X_Antalp = df_train_kpi_antalp.drop('Antal produktionsmedarbejdere (t)',axis = 1)
rf = RandomForestRegressor(n_estimators = 1000)
rf.fit(X_Antalp, Y_Antalp)
sorted_idx = rf.feature_importances_.argsort()
plt.barh(X_Antalp.columns[sorted_idx], rf.feature_importances_[sorted_idx])
plt.xlabel("Random Forest Feature Importance")
plt.show()

Valgte kpi'er:
Antal produktionsmedarbejdere (t-1)
Måned (t)

df_train_antalp_variables = df_train_kpi[['Antal produktionsmedarbejdere (t)','Antal produktionsmedarbejdere (t-1)',
                                          'Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)',
                                          'August (t)','September (t)','Oktober (t)','November (t)','December (t)']]

df_val_antalp_variables1 = pd.DataFrame(df_train[['Antal produktionsmedarbejdere (t)']])
df_val_antalp_variables1 = pd.DataFrame(df_val_antalp_variables1.iloc[-1]).transpose()
df_val_antalp_variables2 = pd.DataFrame(df_val[['Antal produktionsmedarbejdere (t)']])
df_val_antalp_variables = df_val_antalp_variables1.append(df_val_antalp_variables2)
df_val_antalp_variables['Antal produktionsmedarbejdere (t-1)'] = df_val_antalp_variables['Antal produktionsmedarbejdere (t)'].shift(1)
df_val_antalp_variables = df_val_antalp_variables.dropna(axis = 0)
df_val_antalp_variables = df_val_antalp_variables.merge(df_monthly_dummies, left_index = True,right_index = True)

df_test_antalp_variables1 = pd.DataFrame(df_val[['Antal produktionsmedarbejdere (t)']])
df_test_antalp_variables1 = pd.DataFrame(df_test_antalp_variables1.iloc[-1]).transpose()
df_test_antalp_variables2 = pd.DataFrame(df_test[['Antal produktionsmedarbejdere (t)']])
df_test_antalp_variables = df_test_antalp_variables1.append(df_test_antalp_variables2)
df_test_antalp_variables['Antal produktionsmedarbejdere (t-1)'] = df_test_antalp_variables['Antal produktionsmedarbejdere (t)'].shift(1)
df_test_antalp_variables = df_test_antalp_variables.dropna(axis = 0)
df_test_antalp_variables = df_test_antalp_variables.merge(df_monthly_dummies, left_index = True,right_index = True)

Model selection and training

from pycaret.regression import *
from sklearn.metrics import mean_squared_error

Omsætning

setup_omsætning = setup(data = df_train_omsætning_variables, 
                        train_size = 1.0,  
                        test_data = df_val_omsætning_variables,
                        target = 'Omsætning (t)', 
                        fold_strategy = 'timeseries', 
                        normalize = False,
                        transform_target = False,
                        session_id=123)

	Description	Value
0	session_id	123
1	Target	Omsætning (t)
2	Original Data	(46, 5)
3	Missing Values	False
4	Numeric Features	4
5	Categorical Features	0
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 4)
10	Transformed Test Set	(8, 4)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	58ed
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
lr	Linear Regression	0.0631	0.0073	0.0761	0.3176	0.0517	0.1714	0.6100
lar	Least Angle Regression	0.0631	0.0073	0.0761	0.3176	0.0517	0.1714	0.0080
ard	Automatic Relevance Determination	0.0656	0.0076	0.0781	0.1600	0.0526	0.1796	0.0080
br	Bayesian Ridge	0.0659	0.0083	0.0796	0.3050	0.0542	0.1789	0.0080
huber	Huber Regressor	0.0668	0.0079	0.0799	0.1509	0.0538	0.1832	0.0120
tr	TheilSen Regressor	0.0689	0.0087	0.0828	0.1639	0.0577	0.1536	0.2150
omp	Orthogonal Matching Pursuit	0.0778	0.0098	0.0881	-0.2745	0.0593	0.2212	0.0070
et	Extra Trees Regressor	0.0868	0.0124	0.1029	0.1103	0.0711	0.1997	0.0850
ransac	Random Sample Consensus	0.0873	0.0117	0.1034	-0.4605	0.0691	0.2044	0.0100
par	Passive Aggressive Regressor	0.1061	0.0185	0.1180	-0.5381	0.0793	0.2794	0.0130
xgboost	Extreme Gradient Boosting	0.1063	0.0161	0.1190	-1.7753	0.0827	0.2426	0.0700
ada	AdaBoost Regressor	0.0991	0.0163	0.1193	-0.6517	0.0825	0.2704	0.0380
rf	Random Forest Regressor	0.0978	0.0168	0.1199	-0.2297	0.0817	0.2531	0.1080
kr	Kernel Ridge	0.1023	0.0174	0.1221	-0.3646	0.0825	0.2336	0.0100
ridge	Ridge Regression	0.1034	0.0184	0.1250	0.0558	0.0842	0.2576	0.0100
gbr	Gradient Boosting Regressor	0.1040	0.0188	0.1261	-1.9228	0.0841	0.2322	0.0190
catboost	CatBoost Regressor	0.0980	0.0196	0.1287	-0.4958	0.0868	0.1994	0.3070
svm	Support Vector Regression	0.1130	0.0241	0.1344	-0.2499	0.0915	0.2950	0.0080
mlp	MLP Regressor	0.1160	0.0274	0.1480	-0.2847	0.1006	0.3078	0.0240
dt	Decision Tree Regressor	0.1284	0.0280	0.1584	-2.8811	0.1064	0.3215	0.0080
llar	Lasso Least Angle Regression	0.1612	0.0455	0.1966	-0.8035	0.1324	0.5287	0.0080
lasso	Lasso Regression	0.1612	0.0455	0.1966	-0.8035	0.1324	0.5287	0.0080
en	Elastic Net	0.1612	0.0455	0.1966	-0.8035	0.1324	0.5287	0.0100
lightgbm	Light Gradient Boosting Machine	0.1636	0.0464	0.1991	-0.8356	0.1350	0.5367	0.1530

['Linear Regression',
 'Least Angle Regression',
 'Automatic Relevance Determination',
 'Bayesian Ridge',
 'Huber Regressor',
 'TheilSen Regressor',
 'Orthogonal Matching Pursuit',
 'Extra Trees Regressor',
 'Random Sample Consensus',
 'Passive Aggressive Regressor',
 'Extreme Gradient Boosting',
 'AdaBoost Regressor',
 'Random Forest Regressor',
 'Kernel Ridge',
 'Ridge Regression',
 'Gradient Boosting Regressor',
 'CatBoost Regressor',
 'Support Vector Regression',
 'MLP Regressor',
 'Decision Tree Regressor',
 'Lasso Least Angle Regression',
 'Lasso Regression',
 'Elastic Net',
 'Light Gradient Boosting Machine']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_omsætning_variables)
    actual = tuned_model_predictions['Omsætning (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_omsætning_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.2234	0.0545	0.2334	-0.5862	0.1667	0.8114
1	0.1175	0.0164	0.1282	-5.2716	0.0829	0.1955
2	0.0347	0.0018	0.0429	-1.1431	0.0292	0.0777
3	0.1848	0.0579	0.2406	-0.3490	0.1440	0.2745
4	0.1476	0.0671	0.2590	-0.4484	0.2086	0.0515
5	0.1632	0.0271	0.1647	0.4036	0.1124	0.4193
6	0.0820	0.0082	0.0904	-0.0026	0.0602	0.1640
7	0.1303	0.0503	0.2242	0.1099	0.1273	0.1526
8	0.2440	0.0716	0.2676	0.3381	0.1901	2.1944
9	0.1542	0.0259	0.1608	0.0785	0.1030	0.2865
Mean	0.1482	0.0381	0.1812	-0.6871	0.1225	0.4627
SD	0.0591	0.0239	0.0725	1.5918	0.0539	0.6134

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_omsætning = tune_model(best_model)
save_model(final_model_omsætning, 'Omsætning')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1385	0.0285	0.1688	0.1705	0.1226	0.5236
1	0.0831	0.0070	0.0839	-1.6893	0.0550	0.1447
2	0.0296	0.0016	0.0403	-0.8858	0.0268	0.0615
3	0.0819	0.0104	0.1021	0.7570	0.0578	0.1187
4	0.0936	0.0137	0.1172	0.7035	0.1007	0.1142
5	0.0282	0.0009	0.0296	0.9808	0.0190	0.0628
6	0.0328	0.0013	0.0361	0.8396	0.0239	0.0651
7	0.0728	0.0064	0.0797	0.8874	0.0506	0.1363
8	0.0335	0.0014	0.0369	0.9874	0.0229	0.3254
9	0.1289	0.0254	0.1593	0.0962	0.1086	0.2558
Mean	0.0723	0.0097	0.0854	0.2847	0.0588	0.1808
SD	0.0388	0.0096	0.0486	0.8581	0.0367	0.1404

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=[],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Omsætning (t)',
                                       time_features=[])),
                 ('imputer',
                  Simple_Imputer(categorical_strategy='not_available',
                                 fill_value_categorical=None,
                                 fill_value_numerical=None,
                                 numeric_str...
                               batch_size='auto', beta_1=0.9, beta_2=0.999,
                               early_stopping=False, epsilon=1e-08,
                               hidden_layer_sizes=[100, 100, 50],
                               learning_rate='adaptive',
                               learning_rate_init=0.001, max_fun=15000,
                               max_iter=500, momentum=0.9, n_iter_no_change=10,
                               nesterovs_momentum=True, power_t=0.5,
                               random_state=123, shuffle=True, solver='adam',
                               tol=0.0001, validation_fraction=0.1,
                               verbose=False, warm_start=False)]],
          verbose=False),
 'Omsætning.pkl')

rmse.index[0]

'mlp'

Vareforbrug

setup_vareforbrug = setup(data = df_train_vareforbrug_variables, 
                          train_size = 1.0,  
                          test_data = df_val_vareforbrug_variables,
                          target = 'Vareforbrug (t)', 
                          fold_strategy = 'timeseries', 
                          normalize = False,
                          transform_target = False,
                          session_id=123)

	Description	Value
0	session_id	123
1	Target	Vareforbrug (t)
2	Original Data	(46, 6)
3	Missing Values	False
4	Numeric Features	5
5	Categorical Features	0
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 5)
10	Transformed Test Set	(8, 5)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	0e32
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
ard	Automatic Relevance Determination	0.1390	0.0334	0.1698	-0.1349	0.1129	0.2999	0.0080
omp	Orthogonal Matching Pursuit	0.1415	0.0342	0.1708	-0.2486	0.1137	0.2990	0.0090
ridge	Ridge Regression	0.1429	0.0342	0.1731	-0.2249	0.1151	0.3053	0.0100
br	Bayesian Ridge	0.1408	0.0354	0.1748	-0.1110	0.1164	0.3079	0.0070
lightgbm	Light Gradient Boosting Machine	0.1436	0.0379	0.1779	-0.1185	0.1188	0.3009	0.0130
et	Extra Trees Regressor	0.1443	0.0383	0.1782	-0.2352	0.1168	0.3047	0.1010
rf	Random Forest Regressor	0.1407	0.0388	0.1818	-0.4454	0.1187	0.2992	0.1110
ada	AdaBoost Regressor	0.1440	0.0389	0.1821	-0.2674	0.1190	0.3057	0.0500
gbr	Gradient Boosting Regressor	0.1461	0.0400	0.1857	-0.4978	0.1200	0.3003	0.0220
lasso	Lasso Regression	0.1555	0.0412	0.1859	-0.1728	0.1236	0.3265	0.0110
llar	Lasso Least Angle Regression	0.1555	0.0412	0.1859	-0.1728	0.1236	0.3265	0.0070
en	Elastic Net	0.1555	0.0412	0.1859	-0.1728	0.1236	0.3265	0.0090
catboost	CatBoost Regressor	0.1477	0.0407	0.1870	-0.3820	0.1218	0.3207	0.2940
mlp	MLP Regressor	0.1682	0.0416	0.1931	-0.9683	0.1288	0.3579	0.0240
xgboost	Extreme Gradient Boosting	0.1648	0.0464	0.1932	-0.5198	0.1255	0.3526	0.0710
huber	Huber Regressor	0.1626	0.0454	0.1986	-2.6686	0.1287	0.3336	0.0120
svm	Support Vector Regression	0.1620	0.0467	0.2009	-0.6226	0.1317	0.3418	0.0130
dt	Decision Tree Regressor	0.1747	0.0516	0.2109	-2.1022	0.1381	0.3479	0.0080
kr	Kernel Ridge	0.1925	0.0535	0.2235	-2.0386	0.1506	0.3877	0.0070
par	Passive Aggressive Regressor	0.2045	0.0690	0.2532	-4.4142	0.1681	0.4297	0.0080
lr	Linear Regression	0.2083	0.1530	0.2721	-7.7946	0.1531	0.4122	0.0110
tr	TheilSen Regressor	0.2167	0.1566	0.2790	-8.1374	0.1563	0.4181	0.2660
ransac	Random Sample Consensus	0.2704	0.1858	0.3389	-12.1230	0.1959	0.5481	0.0300
lar	Least Angle Regression	0.8685	12.1656	1.2814	-668.8928	0.2743	1.5460	0.0080

['Automatic Relevance Determination',
 'Orthogonal Matching Pursuit',
 'Ridge Regression',
 'Bayesian Ridge',
 'Light Gradient Boosting Machine',
 'Extra Trees Regressor',
 'Random Forest Regressor',
 'AdaBoost Regressor',
 'Gradient Boosting Regressor',
 'Lasso Regression',
 'Lasso Least Angle Regression',
 'Elastic Net',
 'CatBoost Regressor',
 'MLP Regressor',
 'Extreme Gradient Boosting',
 'Huber Regressor',
 'Support Vector Regression',
 'Decision Tree Regressor',
 'Kernel Ridge',
 'Passive Aggressive Regressor',
 'Linear Regression',
 'TheilSen Regressor',
 'Random Sample Consensus',
 'Least Angle Regression']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_vareforbrug_variables)
    actual = tuned_model_predictions['Vareforbrug (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_vareforbrug_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.2530	0.0675	0.2598	-2.7026	0.1783	0.5337
1	0.2249	0.0709	0.2663	-16.6173	0.1891	0.3747
2	0.1259	0.0236	0.1538	-2.4637	0.1014	0.2144
3	0.2358	0.0783	0.2798	-0.6508	0.1712	0.3815
4	0.1244	0.0272	0.1649	0.0701	0.1207	0.4845
5	0.0771	0.0116	0.1075	0.6427	0.0758	0.2197
6	0.2046	0.0469	0.2165	-4.5564	0.1397	0.4305
7	0.2783	0.1070	0.3271	-0.9991	0.1950	0.4358
8	0.2187	0.0783	0.2799	0.3577	0.2128	0.1663
9	0.2079	0.0701	0.2647	-0.1627	0.1820	0.7563
Mean	0.1951	0.0581	0.2320	-2.7082	0.1566	0.3997
SD	0.0611	0.0284	0.0656	4.8880	0.0426	0.1663

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_vareforbrug = tune_model(best_model)
save_model(final_model_vareforbrug, 'Vareforbrug')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1701	0.0486	0.2205	-1.6674	0.1378	0.3700
1	0.1848	0.0627	0.2505	-14.5877	0.1607	0.3323
2	0.1115	0.0233	0.1528	-2.4192	0.0957	0.2278
3	0.1651	0.0347	0.1863	0.2676	0.1151	0.2969
4	0.2073	0.0802	0.2832	-1.7436	0.1962	0.8666
5	0.3721	0.2322	0.4818	-6.1808	0.2493	0.6674
6	0.1193	0.0200	0.1413	-1.3657	0.0932	0.2585
7	0.1683	0.0535	0.2314	-0.0005	0.1345	0.2428
8	0.2777	0.1044	0.3232	0.1438	0.2259	0.2959
9	0.1080	0.0281	0.1676	0.5340	0.1239	0.4840
Mean	0.1884	0.0688	0.2439	-2.7019	0.1532	0.4042
SD	0.0778	0.0600	0.0965	4.3808	0.0512	0.1993

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=[],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Vareforbrug (t)',
                                       time_features=[])),
                 ('imputer',
                  Simple_Imputer(categorical_strategy='not_available',
                                 fill_value_categorical=None,
                                 fill_value_numerical=None,
                                 numeric_s...
                 ('feature_select', 'passthrough'), ('fix_multi', 'passthrough'),
                 ('dfs', 'passthrough'), ('pca', 'passthrough'),
                 ['trained_model',
                  RANSACRegressor(base_estimator=None, is_data_valid=None,
                                  is_model_valid=None, loss='squared_loss',
                                  max_skips=13, max_trials=17, min_samples=0.2,
                                  random_state=123, residual_threshold=None,
                                  stop_n_inliers=10, stop_probability=0.75,
                                  stop_score=inf)]],
          verbose=False),
 'Vareforbrug.pkl')

rmse.index[0]

'ransac'

Lønforbrug

setup_lønforbrug = setup(data = df_train_lønforbrug_variables, 
                         train_size = 1.0,  
                         test_data = df_val_lønforbrug_variables,
                         target = 'Lønforbrug (t)', 
                         fold_strategy = 'timeseries', 
                         normalize = False,
                         transform_target = False,
                         session_id=123)

	Description	Value
0	session_id	123
1	Target	Lønforbrug (t)
2	Original Data	(46, 5)
3	Missing Values	False
4	Numeric Features	4
5	Categorical Features	0
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 4)
10	Transformed Test Set	(8, 4)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	ac31
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
tr	TheilSen Regressor	0.1514	0.0371	0.1730	0.0017	0.1177	0.3877	0.2330
ridge	Ridge Regression	0.1554	0.0350	0.1780	-0.0870	0.1225	0.3989	0.0120
lr	Linear Regression	0.1579	0.0377	0.1783	-0.1948	0.1215	0.3963	0.0090
lar	Least Angle Regression	0.1579	0.0377	0.1783	-0.1948	0.1215	0.3963	0.0120
rf	Random Forest Regressor	0.1543	0.0354	0.1796	-0.1719	0.1236	0.4021	0.1250
br	Bayesian Ridge	0.1562	0.0371	0.1798	-0.1845	0.1231	0.3991	0.0110
ard	Automatic Relevance Determination	0.1616	0.0377	0.1816	-0.2787	0.1240	0.4083	0.0130
huber	Huber Regressor	0.1611	0.0405	0.1828	-0.2853	0.1237	0.4017	0.0110
et	Extra Trees Regressor	0.1585	0.0374	0.1832	-0.3608	0.1252	0.4021	0.1180
ransac	Random Sample Consensus	0.1570	0.0385	0.1841	-0.0703	0.1232	0.3799	0.0160
lightgbm	Light Gradient Boosting Machine	0.1573	0.0379	0.1865	-0.0655	0.1274	0.3889	0.0120
omp	Orthogonal Matching Pursuit	0.1680	0.0406	0.1884	-0.3433	0.1282	0.3985	0.0110
catboost	CatBoost Regressor	0.1536	0.0406	0.1890	-0.3595	0.1288	0.3969	0.3720
en	Elastic Net	0.1614	0.0394	0.1909	-0.1059	0.1305	0.3998	0.0110
llar	Lasso Least Angle Regression	0.1614	0.0394	0.1909	-0.1059	0.1305	0.3998	0.0110
lasso	Lasso Regression	0.1614	0.0394	0.1909	-0.1059	0.1305	0.3998	0.0110
gbr	Gradient Boosting Regressor	0.1680	0.0446	0.1931	-0.4270	0.1303	0.4133	0.0210
svm	Support Vector Regression	0.1780	0.0458	0.2057	-0.4611	0.1397	0.4488	0.0150
ada	AdaBoost Regressor	0.1722	0.0480	0.2066	-0.9043	0.1400	0.4687	0.0410
xgboost	Extreme Gradient Boosting	0.1824	0.0510	0.2109	-0.7021	0.1431	0.4343	0.0940
kr	Kernel Ridge	0.1855	0.0487	0.2123	-0.7150	0.1434	0.4040	0.0090
mlp	MLP Regressor	0.1772	0.0500	0.2152	-0.6603	0.1476	0.4713	0.0190
dt	Decision Tree Regressor	0.2018	0.0686	0.2477	-1.3639	0.1623	0.5146	0.0080
par	Passive Aggressive Regressor	0.2630	0.1055	0.2931	-1.6873	0.1862	0.5928	0.0080

['TheilSen Regressor',
 'Ridge Regression',
 'Linear Regression',
 'Least Angle Regression',
 'Random Forest Regressor',
 'Bayesian Ridge',
 'Automatic Relevance Determination',
 'Huber Regressor',
 'Extra Trees Regressor',
 'Random Sample Consensus',
 'Light Gradient Boosting Machine',
 'Orthogonal Matching Pursuit',
 'CatBoost Regressor',
 'Elastic Net',
 'Lasso Least Angle Regression',
 'Lasso Regression',
 'Gradient Boosting Regressor',
 'Support Vector Regression',
 'AdaBoost Regressor',
 'Extreme Gradient Boosting',
 'Kernel Ridge',
 'MLP Regressor',
 'Decision Tree Regressor',
 'Passive Aggressive Regressor']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_lønforbrug_variables)
    actual = tuned_model_predictions['Lønforbrug (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_lønforbrug_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1708	0.0374	0.1934	-0.2829	0.1279	0.3904
1	0.2788	0.0941	0.3067	-0.3187	0.1974	0.6588
2	0.1362	0.0218	0.1478	-1.8774	0.0990	0.3058
3	0.1863	0.0509	0.2257	0.1190	0.1648	1.0033
4	0.2151	0.0792	0.2814	-0.1736	0.2201	0.1889
5	0.1169	0.0187	0.1369	0.5177	0.0837	0.2066
6	0.1404	0.0249	0.1578	-0.0545	0.1034	0.2908
7	0.1181	0.0158	0.1255	0.2235	0.0858	0.2691
8	0.1349	0.0220	0.1482	-0.3465	0.0960	0.2263
9	0.1824	0.0354	0.1881	0.0408	0.1195	0.3720
Mean	0.1680	0.0400	0.1911	-0.2152	0.1298	0.3912
SD	0.0480	0.0256	0.0590	0.6107	0.0457	0.2411

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_lønforbrug = tune_model(best_model)
save_model(final_model_lønforbrug, 'Lønforbrug')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1490	0.0313	0.1768	-0.0718	0.1212	0.4167
1	0.2463	0.0767	0.2769	-0.0744	0.1816	0.6307
2	0.0929	0.0137	0.1172	-0.8086	0.0792	0.2255
3	0.1628	0.0377	0.1942	0.3477	0.1442	0.8934
4	0.1921	0.0813	0.2851	-0.2047	0.2242	0.1224
5	0.1402	0.0237	0.1539	0.3904	0.1008	0.3096
6	0.1234	0.0225	0.1499	0.0483	0.0983	0.2614
7	0.1346	0.0202	0.1421	0.0045	0.0966	0.3071
8	0.1018	0.0158	0.1257	0.0307	0.0823	0.2009
9	0.1192	0.0229	0.1512	0.3802	0.0979	0.2598
Mean	0.1462	0.0346	0.1773	0.0042	0.1226	0.3627
SD	0.0432	0.0232	0.0560	0.3358	0.0449	0.2204

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=[],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Lønforbrug (t)',
                                       time_features=[])),
                 ('imputer',
                  Simple_Imputer(categorical_strategy='not_available',
                                 fill_value_categorical=None,
                                 fill_value_numerical=None,
                                 numeric_st...
                 ('cluster_all', 'passthrough'),
                 ('dummy', Dummify(target='Lønforbrug (t)')),
                 ('fix_perfect', Remove_100(target='Lønforbrug (t)')),
                 ('clean_names', Clean_Colum_Names()),
                 ('feature_select', 'passthrough'), ('fix_multi', 'passthrough'),
                 ('dfs', 'passthrough'), ('pca', 'passthrough'),
                 ['trained_model',
                  <catboost.core.CatBoostRegressor object at 0x0000024D28320DC0>]],
          verbose=False),
 'Lønforbrug.pkl')

rmse.index[0]

'catboost'

Faktureringsgrad

setup_faktureringsgrad = setup(data = df_train_faktureringsgrad_variables, 
                               train_size = 1.0,  
                               test_data = df_val_faktureringsgrad_variables,
                               target = 'Faktureringsgrad (t)', 
                               fold_strategy = 'timeseries', 
                               categorical_features = ['Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)',
                               'August (t)','September (t)','Oktober (t)','November (t)','December (t)'],
                               normalize = False,
                               transform_target = False,
                               session_id=123)

	Description	Value
0	session_id	123
1	Target	Faktureringsgrad (t)
2	Original Data	(46, 14)
3	Missing Values	False
4	Numeric Features	2
5	Categorical Features	11
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 13)
10	Transformed Test Set	(8, 13)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	9f0f
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
ransac	Random Sample Consensus	0.1670	0.0504	0.1908	-2.7527	0.1178	0.2515	0.0270
kr	Kernel Ridge	0.1654	0.0431	0.1942	-1.5714	0.1230	0.2683	0.0080
ridge	Ridge Regression	0.1750	0.0470	0.2024	-2.1946	0.1282	0.2896	0.0110
br	Bayesian Ridge	0.1761	0.0461	0.2031	-2.0792	0.1301	0.2944	0.0060
par	Passive Aggressive Regressor	0.1769	0.0472	0.2054	-1.7894	0.1308	0.3014	0.0120
ada	AdaBoost Regressor	0.1816	0.0472	0.2072	-2.1887	0.1295	0.2919	0.0370
svm	Support Vector Regression	0.1790	0.0496	0.2079	-2.3125	0.1333	0.2975	0.0090
tr	TheilSen Regressor	0.1769	0.0498	0.2096	-2.3480	0.1337	0.2955	0.3390
ard	Automatic Relevance Determination	0.1828	0.0492	0.2108	-2.4848	0.1363	0.3138	0.0080
catboost	CatBoost Regressor	0.1844	0.0517	0.2116	-2.9139	0.1329	0.3083	0.3080
mlp	MLP Regressor	0.1802	0.0527	0.2117	-1.9622	0.1326	0.3257	0.0180
rf	Random Forest Regressor	0.1884	0.0519	0.2165	-2.6067	0.1359	0.3152	0.1130
lasso	Lasso Regression	0.1909	0.0563	0.2209	-3.0924	0.1381	0.3121	0.0140
en	Elastic Net	0.1909	0.0563	0.2209	-3.0924	0.1381	0.3121	0.0080
llar	Lasso Least Angle Regression	0.1909	0.0563	0.2209	-3.0924	0.1381	0.3121	0.0120
huber	Huber Regressor	0.1943	0.0557	0.2255	-2.3727	0.1474	0.3312	0.0160
lightgbm	Light Gradient Boosting Machine	0.1945	0.0578	0.2258	-3.1851	0.1411	0.3172	0.0150
et	Extra Trees Regressor	0.2056	0.0601	0.2349	-3.0467	0.1520	0.3422	0.1010
xgboost	Extreme Gradient Boosting	0.2081	0.0664	0.2383	-4.1590	0.1515	0.3303	0.0760
lr	Linear Regression	0.2091	0.0662	0.2407	-3.0597	0.1563	0.3515	0.0100
gbr	Gradient Boosting Regressor	0.2121	0.0650	0.2410	-4.0341	0.1518	0.3528	0.0210
omp	Orthogonal Matching Pursuit	0.2128	0.0673	0.2422	-3.4288	0.1557	0.3591	0.0080
dt	Decision Tree Regressor	0.2284	0.0788	0.2593	-5.3512	0.1635	0.3669	0.0080
lar	Least Angle Regression	0.6394	1.1481	0.7346	-96.3225	0.2841	0.9973	0.0100

['Random Sample Consensus',
 'Kernel Ridge',
 'Ridge Regression',
 'Bayesian Ridge',
 'Passive Aggressive Regressor',
 'AdaBoost Regressor',
 'Support Vector Regression',
 'TheilSen Regressor',
 'Automatic Relevance Determination',
 'CatBoost Regressor',
 'MLP Regressor',
 'Random Forest Regressor',
 'Lasso Regression',
 'Elastic Net',
 'Lasso Least Angle Regression',
 'Huber Regressor',
 'Light Gradient Boosting Machine',
 'Extra Trees Regressor',
 'Extreme Gradient Boosting',
 'Linear Regression',
 'Gradient Boosting Regressor',
 'Orthogonal Matching Pursuit',
 'Decision Tree Regressor',
 'Least Angle Regression']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_faktureringsgrad_variables)
    actual = tuned_model_predictions['Faktureringsgrad (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_faktureringsgrad_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.3712	0.1480	0.3847	-14.2704	0.2660	1.1661
1	0.3175	0.1590	0.3987	-0.4274	0.2912	0.3704
2	0.1939	0.0435	0.2086	-4.2951	0.1184	0.2207
3	0.3057	0.1329	0.3646	-7.1399	0.2715	0.5266
4	0.2665	0.1569	0.3962	-6.5121	0.1938	0.3585
5	0.2096	0.0463	0.2151	-2.1093	0.1249	0.3037
6	0.4370	0.2064	0.4543	-5.4568	0.2762	0.7936
7	0.1113	0.0139	0.1178	0.1344	0.0756	0.1943
8	0.1872	0.0539	0.2321	-5.2235	0.1289	0.2167
9	0.4440	0.3203	0.5659	-19.0275	0.2896	0.6928
Mean	0.2844	0.1281	0.3338	-6.4328	0.2036	0.4843
SD	0.1055	0.0881	0.1291	5.7047	0.0802	0.2987

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_faktureringsgrad = tune_model(best_model)
save_model(final_model_faktureringsgrad, 'Faktureringsgrad')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.3072	0.1041	0.3226	-9.7355	0.2163	0.9790
1	0.2674	0.1114	0.3338	-0.0005	0.2380	0.2668
2	0.1274	0.0245	0.1564	-1.9747	0.0839	0.1338
3	0.2578	0.0810	0.2846	-3.9599	0.1823	0.6159
4	0.1268	0.0217	0.1473	-0.0391	0.0813	0.1591
5	0.1158	0.0163	0.1278	-0.0987	0.0737	0.1657
6	0.1815	0.0373	0.1931	-0.1663	0.1179	0.3192
7	0.1879	0.0483	0.2198	-2.0126	0.1368	0.3214
8	0.0925	0.0164	0.1280	-0.8931	0.0706	0.1041
9	0.1470	0.0223	0.1494	-0.3956	0.0896	0.2402
Mean	0.1811	0.0483	0.2063	-1.9276	0.1290	0.3305
SD	0.0696	0.0350	0.0760	2.8679	0.0590	0.2571

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=['Februar (t)',
                                                             'Marts (t)',
                                                             'April (t)',
                                                             'Maj (t)',
                                                             'Juni (t)',
                                                             'Juli (t)',
                                                             'August (t)',
                                                             'September (t)',
                                                             'Oktober (t)',
                                                             'November (t)',
                                                             'December (t)'],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Faktureringsgrad (t)',
                                       time_feat...
                 ('dfs', 'passthrough'), ('pca', 'passthrough'),
                 ['trained_model',
                  DecisionTreeRegressor(ccp_alpha=0.0, criterion='mae',
                                        max_depth=4, max_features='log2',
                                        max_leaf_nodes=None,
                                        min_impurity_decrease=0.0002,
                                        min_impurity_split=None,
                                        min_samples_leaf=5, min_samples_split=5,
                                        min_weight_fraction_leaf=0.0,
                                        presort='deprecated', random_state=123,
                                        splitter='best')]],
          verbose=False),
 'Faktureringsgrad.pkl')

rmse.index[0]

'dt'

Fakturerbar tid

setup_fakturerbartid = setup(data = df_train_fakturerbartid_variables, 
                             train_size = 1.0,  
                             test_data = df_val_fakturerbartid_variables,
                             target = 'Fakturerbar tid (t)', 
                             fold_strategy = 'timeseries', 
                             normalize = False,
                             transform_target = False,
                             session_id=123)

	Description	Value
0	session_id	123
1	Target	Fakturerbar tid (t)
2	Original Data	(46, 4)
3	Missing Values	False
4	Numeric Features	3
5	Categorical Features	0
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 3)
10	Transformed Test Set	(8, 3)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	e989
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
ransac	Random Sample Consensus	0.1383	0.0323	0.1632	-0.4045	0.1090	0.3887	0.0190
huber	Huber Regressor	0.1514	0.0385	0.1834	-0.4718	0.1223	0.4410	0.0130
tr	TheilSen Regressor	0.1496	0.0403	0.1851	-0.3832	0.1234	0.4410	0.2200
kr	Kernel Ridge	0.1565	0.0431	0.1851	-0.2930	0.1250	0.4326	0.0100
ridge	Ridge Regression	0.1475	0.0382	0.1864	-0.5492	0.1248	0.4632	0.0110
catboost	CatBoost Regressor	0.1567	0.0388	0.1877	-0.6058	0.1270	0.4881	0.2470
en	Elastic Net	0.1461	0.0393	0.1900	-0.5564	0.1280	0.4534	0.0110
llar	Lasso Least Angle Regression	0.1461	0.0393	0.1900	-0.5564	0.1280	0.4534	0.0100
lasso	Lasso Regression	0.1461	0.0393	0.1900	-0.5564	0.1280	0.4534	0.0120
lightgbm	Light Gradient Boosting Machine	0.1474	0.0398	0.1908	-0.5635	0.1288	0.4615	0.0160
br	Bayesian Ridge	0.1542	0.0408	0.1914	-0.6265	0.1277	0.4846	0.0110
ada	AdaBoost Regressor	0.1639	0.0431	0.1924	-0.5704	0.1290	0.5081	0.0420
rf	Random Forest Regressor	0.1599	0.0407	0.1930	-0.6407	0.1299	0.4900	0.1110
mlp	MLP Regressor	0.1663	0.0440	0.1937	-0.5806	0.1307	0.5277	0.0260
ard	Automatic Relevance Determination	0.1623	0.0442	0.1978	-0.6874	0.1318	0.5055	0.0110
gbr	Gradient Boosting Regressor	0.1741	0.0458	0.1987	-0.6692	0.1351	0.5326	0.0300
lr	Linear Regression	0.1637	0.0448	0.1988	-0.8949	0.1320	0.5011	0.0100
lar	Least Angle Regression	0.1637	0.0448	0.1988	-0.8949	0.1320	0.5011	0.0120
et	Extra Trees Regressor	0.1740	0.0436	0.2001	-0.8346	0.1348	0.5135	0.1110
svm	Support Vector Regression	0.1710	0.0479	0.2044	-0.7630	0.1386	0.5607	0.0110
omp	Orthogonal Matching Pursuit	0.1721	0.0485	0.2080	-0.8572	0.1387	0.5301	0.0100
xgboost	Extreme Gradient Boosting	0.1836	0.0507	0.2158	-1.1845	0.1460	0.5611	0.0740
dt	Decision Tree Regressor	0.1917	0.0588	0.2329	-1.8433	0.1589	0.6147	0.0120
par	Passive Aggressive Regressor	0.4215	0.2454	0.4659	-15.1002	0.2395	1.1670	0.0120

['Random Sample Consensus',
 'Huber Regressor',
 'TheilSen Regressor',
 'Kernel Ridge',
 'Ridge Regression',
 'CatBoost Regressor',
 'Elastic Net',
 'Lasso Least Angle Regression',
 'Lasso Regression',
 'Light Gradient Boosting Machine',
 'Bayesian Ridge',
 'AdaBoost Regressor',
 'Random Forest Regressor',
 'MLP Regressor',
 'Automatic Relevance Determination',
 'Gradient Boosting Regressor',
 'Linear Regression',
 'Least Angle Regression',
 'Extra Trees Regressor',
 'Support Vector Regression',
 'Orthogonal Matching Pursuit',
 'Extreme Gradient Boosting',
 'Decision Tree Regressor',
 'Passive Aggressive Regressor']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_fakturerbartid_variables)
    actual = tuned_model_predictions['Fakturerbar tid (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_fakturerbartid_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1548	0.0567	0.2382	-0.4090	0.1830	1.5751
1	0.1844	0.0698	0.2641	-0.4062	0.1686	0.3471
2	0.0894	0.0108	0.1041	-1.1870	0.0729	0.2474
3	0.1457	0.0300	0.1733	-2.3084	0.1251	0.4846
4	0.0993	0.0134	0.1156	-0.3933	0.0837	0.3428
5	0.2389	0.0891	0.2985	-1.8054	0.2017	0.4519
6	0.1253	0.0305	0.1747	-0.1381	0.1053	0.1841
7	0.1028	0.0200	0.1415	-0.2131	0.0896	0.1675
8	0.2161	0.0533	0.2309	-0.3301	0.1616	0.6518
9	0.2583	0.0862	0.2936	-0.1717	0.1790	0.4416
Mean	0.1615	0.0460	0.2034	-0.7362	0.1370	0.4894
SD	0.0574	0.0277	0.0678	0.7257	0.0447	0.3881

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_fakturerbartid = tune_model(best_model)
save_model(final_model_fakturerbartid, 'Fakturerbartid')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.2301	0.0621	0.2492	-0.5414	0.1847	1.5038
1	0.2128	0.0661	0.2571	-0.3328	0.1646	0.4478
2	0.0704	0.0073	0.0853	-0.4680	0.0598	0.1929
3	0.1337	0.0257	0.1604	-1.8355	0.1109	0.4102
4	0.0883	0.0113	0.1062	-0.1763	0.0810	0.3457
5	0.1501	0.0427	0.2066	-0.3447	0.1343	0.2626
6	0.1678	0.0389	0.1973	-0.4523	0.1231	0.2692
7	0.0967	0.0119	0.1092	0.2767	0.0687	0.1747
8	0.1207	0.0192	0.1384	0.5218	0.0906	0.2476
9	0.2526	0.0838	0.2896	-0.1397	0.1760	0.4249
Mean	0.1523	0.0369	0.1799	-0.3492	0.1194	0.4279
SD	0.0594	0.0252	0.0673	0.5912	0.0427	0.3701

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=[],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Fakturerbar tid (t)',
                                       time_features=[])),
                 ('imputer',
                  Simple_Imputer(categorical_strategy='not_available',
                                 fill_value_categorical=None,
                                 fill_value_numerical=None,
                                 numeri...
                 ('feature_select', 'passthrough'), ('fix_multi', 'passthrough'),
                 ('dfs', 'passthrough'), ('pca', 'passthrough'),
                 ['trained_model',
                  RANSACRegressor(base_estimator=None, is_data_valid=None,
                                  is_model_valid=None, loss='squared_loss',
                                  max_skips=17, max_trials=7, min_samples=0.5,
                                  random_state=123, residual_threshold=None,
                                  stop_n_inliers=10, stop_probability=0.57,
                                  stop_score=inf)]],
          verbose=False),
 'Fakturerbartid.pkl')

rmse.index[0]

'ransac'

Antal produktionsmedarbejdere

setup_antalp = setup(data = df_train_antalp_variables, 
                     train_size = 1.0,  
                     test_data = df_val_antalp_variables,
                     target = 'Antal produktionsmedarbejdere (t)', 
                     fold_strategy = 'timeseries', 
                     categorical_features = ['Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)',
                     'August (t)','September (t)','Oktober (t)','November (t)','December (t)'],
                     normalize = False,
                     transform_target = False,
                     session_id=123)

	Description	Value
0	session_id	123
1	Target	Antal produktionsmedarbejdere (t)
2	Original Data	(46, 13)
3	Missing Values	False
4	Numeric Features	1
5	Categorical Features	11
6	Ordinal Features	False
7	High Cardinality Features	False
8	High Cardinality Method	None
9	Transformed Train Set	(46, 12)
10	Transformed Test Set	(8, 12)
11	Shuffle Train-Test	True
12	Stratify Train-Test	False
13	Fold Generator	TimeSeriesSplit
14	Fold Number	10
15	CPU Jobs	-1
16	Use GPU	False
17	Log Experiment	False
18	Experiment Name	reg-default-name
19	USI	8fdd
20	Imputation Type	simple
21	Iterative Imputation Iteration	None
22	Numeric Imputer	mean
23	Iterative Imputation Numeric Model	None
24	Categorical Imputer	constant
25	Iterative Imputation Categorical Model	None
26	Unknown Categoricals Handling	least_frequent
27	Normalize	False
28	Normalize Method	None
29	Transformation	False
30	Transformation Method	None
31	PCA	False
32	PCA Method	None
33	PCA Components	None
34	Ignore Low Variance	False
35	Combine Rare Levels	False
36	Rare Level Threshold	None
37	Numeric Binning	False
38	Remove Outliers	False
39	Outliers Threshold	None
40	Remove Multicollinearity	False
41	Multicollinearity Threshold	None
42	Remove Perfect Collinearity	True
43	Clustering	False
44	Clustering Iteration	None
45	Polynomial Features	False
46	Polynomial Degree	None
47	Trignometry Features	False
48	Polynomial Threshold	None
49	Group Features	False
50	Feature Selection	False
51	Feature Selection Method	classic
52	Features Selection Threshold	None
53	Feature Interaction	False
54	Feature Ratio	False
55	Interaction Threshold	None
56	Transform Target	False
57	Transform Target Method	box-cox

models = compare_models(sort ='RMSE',exclude =['knn'],turbo = False)
results = pull()
results.Model.tolist()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
lightgbm	Light Gradient Boosting Machine	0.2217	0.0708	0.2507	-4.5556	0.1733	1.4629	0.0170
ard	Automatic Relevance Determination	0.2163	0.0732	0.2510	-3.1399	0.1737	1.8074	0.0120
en	Elastic Net	0.2219	0.0711	0.2519	-4.5722	0.1741	1.4665	0.0120
llar	Lasso Least Angle Regression	0.2219	0.0711	0.2519	-4.5722	0.1741	1.4665	0.0110
lasso	Lasso Regression	0.2219	0.0711	0.2519	-4.5722	0.1741	1.4665	0.0140
ada	AdaBoost Regressor	0.2248	0.0732	0.2542	-2.8617	0.1785	1.6564	0.0460
rf	Random Forest Regressor	0.2204	0.0747	0.2547	-3.3659	0.1771	1.5858	0.1210
kr	Kernel Ridge	0.2168	0.0728	0.2583	-3.9231	0.1791	1.4592	0.0140
br	Bayesian Ridge	0.2264	0.0772	0.2583	-4.6648	0.1778	1.7561	0.0110
ridge	Ridge Regression	0.2262	0.0744	0.2587	-3.9795	0.1797	1.6614	0.0140
svm	Support Vector Regression	0.2212	0.0743	0.2589	-4.1594	0.1804	1.6507	0.0100
omp	Orthogonal Matching Pursuit	0.2276	0.0810	0.2657	-3.4147	0.1812	1.5940	0.0110
huber	Huber Regressor	0.2204	0.0794	0.2680	-3.2422	0.1864	1.7112	0.0220
xgboost	Extreme Gradient Boosting	0.2375	0.0862	0.2720	-3.7561	0.1903	1.5490	0.1010
tr	TheilSen Regressor	0.2314	0.0857	0.2779	-4.1533	0.1910	1.7441	0.3190
lr	Linear Regression	0.2362	0.0872	0.2802	-4.1446	0.1922	1.7416	0.0090
mlp	MLP Regressor	0.2353	0.0896	0.2815	-4.8645	0.1943	1.8385	0.0270
catboost	CatBoost Regressor	0.2432	0.0900	0.2853	-4.8871	0.1994	1.6889	0.3640
gbr	Gradient Boosting Regressor	0.2484	0.0941	0.2879	-5.1773	0.2014	1.7542	0.0260
par	Passive Aggressive Regressor	0.2449	0.0924	0.2889	-3.3545	0.2002	1.6755	0.0120
ransac	Random Sample Consensus	0.2457	0.1583	0.2939	-9.8119	0.1802	0.8704	0.0370
dt	Decision Tree Regressor	0.2745	0.1051	0.3051	-5.2172	0.2116	1.8125	0.0110
et	Extra Trees Regressor	0.2678	0.1063	0.3104	-5.8343	0.2149	1.8791	0.1110
lar	Least Angle Regression	1.3187	16.4087	1.5357	-1262.3369	0.3918	5.7304	0.0110

['Light Gradient Boosting Machine',
 'Automatic Relevance Determination',
 'Elastic Net',
 'Lasso Least Angle Regression',
 'Lasso Regression',
 'AdaBoost Regressor',
 'Random Forest Regressor',
 'Kernel Ridge',
 'Bayesian Ridge',
 'Ridge Regression',
 'Support Vector Regression',
 'Orthogonal Matching Pursuit',
 'Huber Regressor',
 'Extreme Gradient Boosting',
 'TheilSen Regressor',
 'Linear Regression',
 'MLP Regressor',
 'CatBoost Regressor',
 'Gradient Boosting Regressor',
 'Passive Aggressive Regressor',
 'Random Sample Consensus',
 'Decision Tree Regressor',
 'Extra Trees Regressor',
 'Least Angle Regression']

rmse = pd.DataFrame()

for i in range (0,len(results)):
    best_model = create_model(results.index[i])
    tuned_model = tune_model(best_model)
    currentmodelname = results.index[i]
    
    tuned_model_predictions = predict_model(tuned_model,data = df_val_antalp_variables)
    actual = tuned_model_predictions['Antal produktionsmedarbejdere (t)']
    predictions = tuned_model_predictions['Label']

    forecast_actual = list(zip(predictions,actual))
    forecast_period = df_val_antalp_variables.index
    forecast_result = pd.DataFrame(forecast_actual, columns = ['Forecast','Faktiske'], index = forecast_period, dtype='float')
    
    
    rmse.loc[currentmodelname,'RMSE'] = mean_squared_error(actual,predictions,squared=False)

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.2291	0.0579	0.2406	-0.9238	0.1552	0.5250
1	0.4329	0.2407	0.4906	-2.8956	0.3680	11.0646
2	0.3133	0.1418	0.3766	-9.9364	0.2431	1.1274
3	0.1961	0.0668	0.2585	-0.6351	0.1947	1.3882
4	0.1726	0.0521	0.2284	-0.7451	0.1680	0.7529
5	0.2239	0.0805	0.2837	-1.0908	0.1931	0.7194
6	1.0170	1.4476	1.2032	-112.2099	0.5087	1.9271
7	0.2520	0.0758	0.2753	-18.3041	0.1633	0.2959
8	0.2751	0.0987	0.3141	-0.0568	0.2184	1.1260
9	0.1015	0.0137	0.1171	0.3107	0.0775	0.1761
Mean	0.3213	0.2276	0.3788	-14.6487	0.2290	1.9103
SD	0.2465	0.4109	0.2900	33.0007	0.1169	3.0918

rmse = rmse.sort_values(by=['RMSE'])

best_model = create_model(rmse.index[0])
final_model_antalp = tune_model(best_model)
save_model(final_model_antalp, 'Antalp')

	MAE	MSE	RMSE	R2	RMSLE	MAPE
0	0.1961	0.0400	0.1999	-0.3283	0.1296	0.4293
1	0.3165	0.1589	0.3986	-1.5723	0.3130	8.7024
2	0.1697	0.0348	0.1867	-1.6868	0.1372	0.6498
3	0.1958	0.0513	0.2264	-0.2546	0.1759	1.4201
4	0.1735	0.0334	0.1827	-0.1173	0.1379	0.7472
5	0.2213	0.0636	0.2523	-0.6528	0.1694	0.5559
6	0.1761	0.0453	0.2128	-2.5426	0.1369	0.2730
7	0.3403	0.1212	0.3482	-29.8895	0.2082	0.4010
8	0.2639	0.0996	0.3155	-0.0665	0.2093	0.9780
9	0.0903	0.0083	0.0911	0.5825	0.0616	0.1953
Mean	0.2143	0.0656	0.2414	-3.6528	0.1679	1.4352
SD	0.0707	0.0442	0.0857	8.7903	0.0633	2.4465

Transformation Pipeline and Model Successfully Saved

(Pipeline(memory=None,
          steps=[('dtypes',
                  DataTypes_Auto_infer(categorical_features=['Februar (t)',
                                                             'Marts (t)',
                                                             'April (t)',
                                                             'Maj (t)',
                                                             'Juni (t)',
                                                             'Juli (t)',
                                                             'August (t)',
                                                             'September (t)',
                                                             'Oktober (t)',
                                                             'November (t)',
                                                             'December (t)'],
                                       display_types=True, features_todrop=[],
                                       id_columns=[], ml_usecase='regression',
                                       numerical_features=[],
                                       target='Antal produktionsmedarbejdere '
                                              '...
                  RandomForestRegressor(bootstrap=True, ccp_alpha=0.0,
                                        criterion='mse', max_depth=6,
                                        max_features='log2', max_leaf_nodes=None,
                                        max_samples=None,
                                        min_impurity_decrease=0.0002,
                                        min_impurity_split=None,
                                        min_samples_leaf=2, min_samples_split=2,
                                        min_weight_fraction_leaf=0.0,
                                        n_estimators=290, n_jobs=-1,
                                        oob_score=False, random_state=123,
                                        verbose=0, warm_start=False)]],
          verbose=False),
 'Antalp.pkl')

rmse.index[0]

'rf'

Prepare for forecast

df_ferie1 = pd.read_excel (r'C:\Users\jupe\Desktop\Praktik\210\210 - Forecast input.xlsx',sheet_name='Ferie andel af året',
                                     skiprows = 2)
df_ferie1 = df_ferie1.drop(['Måned'], axis = 1)
df_ferie2 = pd.read_excel (r'C:\Users\jupe\Desktop\Praktik\210\210 - Forecast input.xlsx',sheet_name='Ferie andel af året',
                                     skiprows = 2)
df_ferie2 = df_ferie2.drop(['Måned'], axis = 1)
df_ferie = df_ferie1.append(df_ferie2)
df_ferie['Periode'] = ['2021 - 01','2021 - 02','2021 - 03','2021 - 04','2021 - 05','2021 - 06','2021 - 07','2021 - 08',
                       '2021 - 09','2021 - 10','2021 - 11','2021 - 12','2022 - 01','2022 - 02','2022 - 03','2022 - 04',
                       '2022 - 05','2022 - 06','2022 - 07','2022 - 08','2022 - 09','2022 - 10','2022 - 11','2022 - 12']
df_ferie.Periode = pd.to_datetime(df_ferie.Periode)
df_ferie = df_ferie.set_index('Periode')

arbejdsdage = {'Arbejdsdage':[21,20,23,22,21,22,22,22,22,21,22,23,21,20,23,21,22,22,21,23,22,21,22,22],
              'Periode':['2021 - 01','2021 - 02','2021 - 03','2021 - 04','2021 - 05','2021 - 06','2021 - 07','2021 - 08',
                         '2021 - 09','2021 - 10','2021 - 11','2021 - 12','2022 - 01','2022 - 02','2022 - 03','2022 - 04',
                         '2022 - 05','2022 - 06','2022 - 07','2022 - 08','2022 - 09','2022 - 10','2022 - 11','2022 - 12']}
df_arbejdsdage = pd.DataFrame(arbejdsdage)
df_arbejdsdage.Periode = pd.to_datetime(df_arbejdsdage.Periode)
df_arbejdsdage = df_arbejdsdage.set_index('Periode')
df_arbejdsdage_ferie_forecast = df_arbejdsdage.merge(df_ferie,left_index = True, right_index = True)
df_arbejdsdage_ferie_forecast['Arbejdsdage - ferie (t)']= df_arbejdsdage_ferie_forecast['Arbejdsdage']-(df_arbejdsdage_ferie_forecast['Gennemsnit af Ferie andel af året']*30)
df_arbejdsdage_ferie_forecast = df_arbejdsdage_ferie_forecast.drop(['Gennemsnit af Ferie andel af året','Arbejdsdage'], axis = 1)
df_arbejdsdage_ferie_forecast = (df_arbejdsdage_ferie_forecast-arbejdsdage_train_min)/(arbejdsdage_train_max-arbejdsdage_train_min)

df_forecast_periode = pd.DataFrame()
df_forecast_periode['Periode'] = ['2021 - 01','2021 - 02','2021 - 03','2021 - 04','2021 - 05','2021 - 06','2021 - 07',
                                  '2021 - 08','2021 - 09','2021 - 10','2021 - 11','2021 - 12','2022 - 01','2022 - 02',
                                  '2022 - 03','2022 - 04','2022 - 05','2022 - 06','2022 - 07','2022 - 08','2022 - 09',
                                  '2022 - 10','2022 - 11','2022 - 12']
df_forecast_periode.Periode = pd.to_datetime(df_forecast_periode.Periode)

omsætning_t_1 = pd.DataFrame(df_val_omsætning_variables[['Omsætning (t-1)']])
omsætning_t_1 = pd.DataFrame(omsætning_t_1.iloc[0]).transpose()
omsætning_t_1_forecast = df_forecast_periode
omsætning_t_1_forecast = omsætning_t_1_forecast.set_index('Periode')
omsætning_t_1_forecast = omsætning_t_1_forecast.merge(omsætning_t_1,how = 'left',left_index = True,right_index = True)

vareforbrug_t_1 = pd.DataFrame(df_val_omsætning_variables[['Vareforbrug (t-1)']])
vareforbrug_t_1 = pd.DataFrame(vareforbrug_t_1.iloc[0]).transpose()
vareforbrug_t_1_forecast = df_forecast_periode
vareforbrug_t_1_forecast = vareforbrug_t_1_forecast.set_index('Periode')
vareforbrug_t_1_forecast = vareforbrug_t_1_forecast.merge(vareforbrug_t_1,how = 'left',left_index = True,right_index = True)

lønforbrug_t_1 = pd.DataFrame(df_val_lønforbrug_variables[['Lønforbrug (t-1)']])
lønforbrug_t_1 = pd.DataFrame(lønforbrug_t_1.iloc[0]).transpose()
lønforbrug_t_1_forecast = df_forecast_periode
lønforbrug_t_1_forecast = lønforbrug_t_1_forecast.set_index('Periode')
lønforbrug_t_1_forecast = lønforbrug_t_1_forecast.merge(lønforbrug_t_1,how = 'left',left_index = True,right_index = True)

faktureringsgrad_t_1 = pd.DataFrame(df_val_faktureringsgrad_variables[['Faktureringsgrad (t-1)']])
faktureringsgrad_t_1 = pd.DataFrame(faktureringsgrad_t_1.iloc[0]).transpose()
faktureringsgrad_t_1_forecast = df_forecast_periode
faktureringsgrad_t_1_forecast = faktureringsgrad_t_1_forecast.set_index('Periode')
faktureringsgrad_t_1_forecast = faktureringsgrad_t_1_forecast.merge(faktureringsgrad_t_1,how = 'left',left_index = True,
                                                                    right_index = True)

fakturerbartid_t_1 = pd.DataFrame(df_val_fakturerbartid_variables[['Fakturerbar tid (t-1)']])
fakturerbartid_t_1 = pd.DataFrame(fakturerbartid_t_1.iloc[0]).transpose()
fakturerbartid_t_1_forecast = df_forecast_periode
fakturerbartid_t_1_forecast = fakturerbartid_t_1_forecast.set_index('Periode')
fakturerbartid_t_1_forecast = fakturerbartid_t_1_forecast.merge(fakturerbartid_t_1,how = 'left',left_index = True,
                                                                    right_index = True)

antalp_t_1 = pd.DataFrame(df_val_antalp_variables[['Antal produktionsmedarbejdere (t-1)']])
antalp_t_1 = pd.DataFrame(antalp_t_1.iloc[0]).transpose()
antalp_t_1_forecast = df_forecast_periode
antalp_t_1_forecast = antalp_t_1_forecast.set_index('Periode')
antalp_t_1_forecast = antalp_t_1_forecast.merge(antalp_t_1,how = 'left',left_index = True,
                                                                    right_index = True)

monthly_forecast_dummies = df_forecast_periode
monthly_forecast_dummies = monthly_forecast_dummies.set_index('Periode')
monthly_forecast_dummies['Måned (t)'] = monthly_forecast_dummies.index.month
dummies = pd.get_dummies(monthly_forecast_dummies['Måned (t)'])
monthly_forecast_dummies = monthly_forecast_dummies.drop('Måned (t)',axis = 1)
dummies = dummies.drop(1,axis = 1)
dummies.columns = ['Februar (t)','Marts (t)','April (t)','Maj (t)','Juni (t)','Juli (t)','August (t)','September (t)',
                   'Oktober (t)','November (t)','December (t)']
monthly_forecast_dummies = monthly_forecast_dummies.join(dummies)

yearly_forecast_variable = df_forecast_periode
yearly_forecast_variable = yearly_forecast_variable.set_index('Periode')
yearly_forecast_variable['År (t)'] = yearly_forecast_variable.index.year

df_forecast_omsætning = pd.DataFrame(columns = ['Omsætning (t)','Vareforbrug (t)','Jobtimer (t)'],
                                     index = df_forecast_periode.Periode)
df_forecast_omsætning['Omsætning (t-1)'] = omsætning_t_1_forecast
df_forecast_omsætning['Vareforbrug (t-1)'] = vareforbrug_t_1_forecast

df_forecast_vareforbrug = pd.DataFrame(columns = ['Vareforbrug (t)','Jobtimer (t)','Antal produktionsmedarbejdere (t)',
                                                  'Lønforbrug (t)'],
                                       index = df_forecast_periode.Periode)
df_forecast_vareforbrug['Omsætning (t-1)'] = omsætning_t_1_forecast
df_forecast_vareforbrug['Vareforbrug (t-1)'] = vareforbrug_t_1_forecast

df_forecast_lønforbrug = pd.DataFrame(columns = ['Lønforbrug (t)','Jobtimer (t)','Antal produktionsmedarbejdere (t)'],
                                      index = df_forecast_periode.Periode)
df_forecast_lønforbrug['Lønforbrug (t-1)'] = lønforbrug_t_1_forecast
df_forecast_lønforbrug['Arbejdsdage - ferie (t)'] = df_arbejdsdage_ferie_forecast

df_forecast_faktureringsgrad = pd.DataFrame(columns = ['Faktureringsgrad (t)'],index = df_forecast_periode.Periode)
df_forecast_faktureringsgrad = df_forecast_faktureringsgrad.merge(monthly_forecast_dummies,left_index = True, 
                                                                  right_index = True)
df_forecast_faktureringsgrad['Faktureringsgrad (t-1)'] = faktureringsgrad_t_1_forecast
df_forecast_faktureringsgrad['Arbejdsdage - ferie (t)'] = df_arbejdsdage_ferie_forecast

df_forecast_fakturerbartid = pd.DataFrame(columns = ['Fakturerbar tid (t)','Antal produktionsmedarbejdere (t)'],
                                          index = df_forecast_periode.Periode)
df_forecast_fakturerbartid['Fakturerbar tid (t-1)'] = fakturerbartid_t_1_forecast
df_forecast_fakturerbartid['Arbejdsdage - ferie (t)'] = df_arbejdsdage_ferie_forecast

df_forecast_antalp = pd.DataFrame(columns = ['Antal produktionsmedarbejdere (t)'],
                                  index = df_forecast_periode.Periode)
df_forecast_antalp['Antal produktionsmedarbejdere (t-1)'] = antalp_t_1_forecast
df_forecast_antalp = df_forecast_antalp.merge(monthly_forecast_dummies,left_index = True,right_index = True)

final_model_omsætning = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Omsætning')
final_model_vareforbrug = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Vareforbrug')
final_model_lønforbrug = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Lønforbrug')
final_model_faktureringsgrad = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Faktureringsgrad')
final_model_fakturerbartid = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Fakturerbartid')
final_model_antalp = load_model(r'C:\Users\jupe\Desktop\Praktik\210\Antalp')

Transformation Pipeline and Model Successfully Loaded
Transformation Pipeline and Model Successfully Loaded
Transformation Pipeline and Model Successfully Loaded
Transformation Pipeline and Model Successfully Loaded
Transformation Pipeline and Model Successfully Loaded
Transformation Pipeline and Model Successfully Loaded

for i in range (0,len(df_forecast_omsætning)):
    
    Måned_år = df_forecast_periode.Periode[i]
    
    if i > 0:
        Sidste_måned_år = df_forecast_periode.Periode[i-1]
        
        sidste_måned_antalp = np.array(df_forecast_antalp.loc[Sidste_måned_år,'Antal produktionsmedarbejdere (t)'])
        sidste_måned_vareforbrug = np.array(df_forecast_vareforbrug.loc[Sidste_måned_år,'Vareforbrug (t)'])
        sidste_måned_omsætning = np.array(df_forecast_omsætning.loc[Sidste_måned_år,'Omsætning (t)'])
        sidste_måned_lønforbrug = np.array(df_forecast_lønforbrug.loc[Sidste_måned_år,'Lønforbrug (t)'])
        sidste_måned_jobtimer = np.array(df_forecast_lønforbrug.loc[Sidste_måned_år,'Jobtimer (t)'])
        sidste_måned_faktureringsgrad = np.array(df_forecast_faktureringsgrad.loc[Sidste_måned_år,'Faktureringsgrad (t)'])  
        sidste_måned_fakturerbartid = np.array(df_forecast_fakturerbartid.loc[Sidste_måned_år,'Fakturerbar tid (t)'])  
        
        df_forecast_omsætning.loc[Måned_år,'Omsætning (t-1)'] = sidste_måned_omsætning
        df_forecast_omsætning.loc[Måned_år,'Vareforbrug (t-1)'] = sidste_måned_vareforbrug
        df_forecast_lønforbrug.loc[Måned_år,'Lønforbrug (t-1)'] = sidste_måned_lønforbrug
        df_forecast_faktureringsgrad.loc[Måned_år,'Faktureringsgrad (t-1)'] = sidste_måned_faktureringsgrad        
        df_forecast_fakturerbartid.loc[Måned_år,'Fakturerbar tid (t-1)'] = sidste_måned_fakturerbartid
        df_forecast_antalp.loc[Måned_år,'Antal produktionsmedarbejdere (t-1)'] = sidste_måned_antalp
            
    predictions_antal_p = predict_model(final_model_antalp, data = df_forecast_antalp)
    predicted_antal_p = predictions_antal_p.iloc[i,-1]
    df_forecast_antalp.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = predicted_antal_p

    df_forecast_fakturerbartid.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = predicted_antal_p
    predictions_fakturerbartid = predict_model(final_model_fakturerbartid, data = df_forecast_fakturerbartid)
    predicted_fakturerbartid = predictions_fakturerbartid.iloc[i,-1]
    df_forecast_fakturerbartid.loc[Måned_år,'Fakturerbar tid (t)'] = predicted_fakturerbartid
    
    predictions_faktureringsgrad = predict_model(final_model_faktureringsgrad, data = df_forecast_faktureringsgrad)
    predicted_faktureringsgrad = predictions_faktureringsgrad.iloc[i,-1]
    df_forecast_faktureringsgrad.loc[Måned_år,'Faktureringsgrad (t)'] = predicted_faktureringsgrad
        
    predicted_faktureringsgrad_unscaled = predicted_faktureringsgrad*(faktureringsgrad_train_max-faktureringsgrad_train_min)+faktureringsgrad_train_min
    
    if i == 0:
        fakturerbartid_t_1 = df['Fakturerbar tid (t)']
        fakturerbartid_t_1 = fakturerbartid_t_1.loc['2020-12-01 00:00:00']
        fakturerbartid_t_1 = np.log(fakturerbartid_t_1)
        fakturerbartid_t_1 = fakturerbartid_t_1-periodic_estimate_fakturerbartid[11:12].values
    else:
        fakturerbartid_t_1 = predicted_fakturerbartid_unscaled
        fakturerbartid_t_1 = np.log(fakturerbartid_t_1)
        fakturerbartid_t_1 = fakturerbartid_t_1-periodic_estimate_fakturerbartid[i-1]
        
    predicted_fakturerbartid_unscaled = predicted_fakturerbartid*(fakturerbartid_train_max-fakturerbartid_train_min)+fakturerbartid_train_min
    predicted_fakturerbartid_unscaled = predicted_fakturerbartid_unscaled+fakturerbartid_t_1
    predicted_fakturerbartid_unscaled = predicted_fakturerbartid_unscaled+periodic_estimate_fakturerbartid[i]
    predicted_fakturerbartid_unscaled = np.exp(predicted_fakturerbartid_unscaled)
    
    if i == 0:
        jobtimer_t_1 = df['Jobtimer (t)']
        jobtimer_t_1 = jobtimer_t_1.loc['2020-12-01 00:00:00']
        jobtimer_t_1 = jobtimer_t_1-periodic_estimate_jobtimer[11:12].values
    else:
        jobtimer_t_1 = predicted_jobtimer_unscaled
        jobtimer_t_1 = jobtimer_t_1-periodic_estimate_jobtimer[i-1]
    
    predicted_jobtimer_unscaled = predicted_fakturerbartid_unscaled/predicted_faktureringsgrad_unscaled
    predicted_jobtimer = predicted_jobtimer_unscaled-periodic_estimate_jobtimer[i]
    predicted_jobtimer = predicted_jobtimer-jobtimer_t_1
    predicted_jobtimer = (predicted_jobtimer-jobtimer_train_min)/(jobtimer_train_max-jobtimer_train_min)
    
    df_forecast_lønforbrug.loc[Måned_år,'Jobtimer (t)'] = predicted_jobtimer
    df_forecast_lønforbrug.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = predicted_antal_p
    predictions_lønforbrug = predict_model(final_model_lønforbrug, data = df_forecast_lønforbrug)
    predicted_lønforbrug = predictions_lønforbrug.iloc[i,-1]
    df_forecast_lønforbrug.loc[Måned_år,'Lønforbrug (t)'] = predicted_lønforbrug
    
    df_forecast_vareforbrug.loc[Måned_år,'Jobtimer (t)'] = predicted_jobtimer
    df_forecast_vareforbrug.loc[Måned_år,'Lønforbrug (t)'] = predicted_lønforbrug
    predictions_vareforbrug = predict_model(final_model_vareforbrug, data = df_forecast_vareforbrug)
    predicted_vareforbrug = predictions_vareforbrug.iloc[i,-1]
    df_forecast_vareforbrug.loc[Måned_år,'Vareforbrug (t)'] = predicted_vareforbrug
   
    df_forecast_omsætning.loc[Måned_år,'Vareforbrug (t)'] = predicted_vareforbrug    
    df_forecast_omsætning.loc[Måned_år,'Jobtimer (t)'] = predicted_jobtimer
    predictions_omsætning = predict_model(final_model_omsætning, data = df_forecast_omsætning)
    predicted_omsætning = predictions_omsætning.iloc[i,-1]
    df_forecast_omsætning.loc[Måned_år,'Omsætning (t)'] = predicted_omsætning

Forecast transformation

df_forecast = pd.DataFrame()
df_forecast['Omsætning (t)'] = df_forecast_omsætning['Omsætning (t)'] 
df_forecast['Vareforbrug (t)'] = df_forecast_vareforbrug['Vareforbrug (t)']
df_forecast['Lønforbrug (t)'] = df_forecast_lønforbrug['Lønforbrug (t)']
df_forecast['Faktureringsgrad (t)'] = df_forecast_faktureringsgrad['Faktureringsgrad (t)']
df_forecast['Fakturerbar tid (t)'] = df_forecast_fakturerbartid['Fakturerbar tid (t)']
df_forecast['Antal produktionsmedarbejdere (t)'] = df_forecast_antalp['Antal produktionsmedarbejdere (t)']
df_forecast['Jobtimer (t)'] = df_forecast_vareforbrug['Jobtimer (t)']

df_forecast['Omsætning (t)'] = df_forecast['Omsætning (t)']*(omsætning_train_max-omsætning_train_min)+omsætning_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        omsætning2020_12 = df['Omsætning (t)']
        omsætning2020_12 = omsætning2020_12.loc['2020-12-01 00:00:00']
        omsætning2020_12 = omsætning2020_12-periodic_estimate_omsætning[11:12].values
        df_forecast.loc[Måned_år,'Omsætning (t)'] = omsætning2020_12+df_forecast['Omsætning (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Omsætning (t)']= df_forecast['Omsætning (t)'].iloc[i]+periodic_estimate_omsætning[i]
    else:
        omsætning_t_1 = df_forecast['Omsætning (t)'].iloc[i-1]-periodic_estimate_omsætning[i-1]
        df_forecast.loc[Måned_år,'Omsætning (t)'] = omsætning_t_1+df_forecast['Omsætning (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Omsætning (t)'] = df_forecast['Omsætning (t)'].iloc[i]+periodic_estimate_omsætning[i]

df_forecast['Vareforbrug (t)'] = df_forecast['Vareforbrug (t)']*(vareforbrug_train_max-vareforbrug_train_min)+vareforbrug_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        vareforbrug2020_12 = df['Vareforbrug (t)']
        vareforbrug2020_12 = vareforbrug2020_12.loc['2020-12-01 00:00:00']
        vareforbrug2020_12 = vareforbrug2020_12-periodic_estimate_vareforbrug[11:12].values
        df_forecast.loc[Måned_år,'Vareforbrug (t)'] = vareforbrug2020_12+df_forecast['Vareforbrug (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Vareforbrug (t)']= df_forecast['Vareforbrug (t)'].iloc[i]+periodic_estimate_vareforbrug[i]
    else:
        vareforbrug_t_1 = df_forecast['Vareforbrug (t)'].iloc[i-1]-periodic_estimate_vareforbrug[i-1]
        df_forecast.loc[Måned_år,'Vareforbrug (t)'] = vareforbrug_t_1+df_forecast['Vareforbrug (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Vareforbrug (t)'] = df_forecast['Vareforbrug (t)'].iloc[i]+periodic_estimate_vareforbrug[i]

df_forecast['Lønforbrug (t)'] = df_forecast['Lønforbrug (t)']*(lønforbrug_train_max-lønforbrug_train_min)+lønforbrug_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        lønforbrug2020_12 = df['Lønforbrug (t)']
        lønforbrug2020_12 = lønforbrug2020_12.loc['2020-12-01 00:00:00']
        lønforbrug2020_12 = lønforbrug2020_12-periodic_estimate_lønforbrug[11:12].values
        df_forecast.loc[Måned_år,'Lønforbrug (t)'] = lønforbrug2020_12+df_forecast['Lønforbrug (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Lønforbrug (t)']= df_forecast['Lønforbrug (t)'].iloc[i]+periodic_estimate_lønforbrug[i]
    else:
        lønforbrug_t_1 = df_forecast['Lønforbrug (t)'].iloc[i-1]-periodic_estimate_lønforbrug[i-1]
        df_forecast.loc[Måned_år,'Lønforbrug (t)'] = lønforbrug_t_1+df_forecast['Lønforbrug (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Lønforbrug (t)'] = df_forecast['Lønforbrug (t)'].iloc[i]+periodic_estimate_lønforbrug[i]

df_forecast['Faktureringsgrad (t)'] = df_forecast['Faktureringsgrad (t)']*(faktureringsgrad_train_max-faktureringsgrad_train_min)+faktureringsgrad_train_min

df_forecast['Fakturerbar tid (t)'] = df_forecast['Fakturerbar tid (t)']*(fakturerbartid_train_max-fakturerbartid_train_min)+fakturerbartid_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        fakturerbartid2020_12 = df['Fakturerbar tid (t)']
        fakturerbartid2020_12 = fakturerbartid2020_12.loc['2020-12-01 00:00:00']
        fakturerbartid2020_12 = np.log(fakturerbartid2020_12)
        fakturerbartid2020_12 = fakturerbartid2020_12-periodic_estimate_fakturerbartid[11:12].values
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)'] = fakturerbartid2020_12+df_forecast['Fakturerbar tid (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)']= df_forecast['Fakturerbar tid (t)'].iloc[i]+periodic_estimate_fakturerbartid[i]
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)']= np.exp(df_forecast['Fakturerbar tid (t)'].iloc[i])
    else:
        fakturerbartid_t_1 = np.log(df_forecast['Fakturerbar tid (t)'].iloc[i-1])
        fakturerbartid_t_1 = fakturerbartid_t_1 -periodic_estimate_fakturerbartid[i-1]
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)'] = fakturerbartid_t_1+df_forecast['Fakturerbar tid (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)'] = df_forecast['Fakturerbar tid (t)'].iloc[i]+periodic_estimate_fakturerbartid[i]
        df_forecast.loc[Måned_år,'Fakturerbar tid (t)']= np.exp(df_forecast['Fakturerbar tid (t)'].iloc[i])

df_forecast['Antal produktionsmedarbejdere (t)'] = df_forecast['Antal produktionsmedarbejdere (t)']*(antalp_train_max-antalp_train_min)+antalp_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        antalp2020_12 = df['Antal produktionsmedarbejdere (t)']
        antalp2020_12 = antalp2020_12.loc['2020-12-01 00:00:00']
        antalp2020_12 = np.log(antalp2020_12)
        antalp2020_12 = antalp2020_12-periodic_estimate_antalp[11:12].values
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = antalp2020_12+df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)']= df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i]+periodic_estimate_antalp[i]
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)']= np.exp(df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i])
    else:
        antalp_t_1 = np.log(df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i-1])
        antalp_t_1 = antalp_t_1-periodic_estimate_antalp[i-1]
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = antalp_t_1+df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)'] = df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i]+periodic_estimate_antalp[i]
        df_forecast.loc[Måned_år,'Antal produktionsmedarbejdere (t)']= np.exp(df_forecast['Antal produktionsmedarbejdere (t)'].iloc[i])

df_forecast['Jobtimer (t)'] = df_forecast['Jobtimer (t)']*(jobtimer_train_max-jobtimer_train_min)+jobtimer_train_min

for i in range (0,len(df_forecast)):
    Måned_år = df_forecast.index[i]
    if i == 0:
        jobtimer2020_12 = df['Jobtimer (t)']
        jobtimer2020_12 = jobtimer2020_12.loc['2020-12-01 00:00:00']
        jobtimer2020_12 = jobtimer2020_12-periodic_estimate_jobtimer[11:12].values
        df_forecast.loc[Måned_år,'Jobtimer (t)'] = jobtimer2020_12+df_forecast['Jobtimer (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Jobtimer (t)']= df_forecast['Jobtimer (t)'].iloc[i]+periodic_estimate_jobtimer[i]
    else:
        jobtimer_t_1 = df_forecast['Jobtimer (t)'].iloc[i-1]-periodic_estimate_jobtimer[i-1]
        df_forecast.loc[Måned_år,'Jobtimer (t)'] = jobtimer_t_1+df_forecast['Jobtimer (t)'].iloc[i]
        df_forecast.loc[Måned_år,'Jobtimer (t)'] = df_forecast['Jobtimer (t)'].iloc[i]+periodic_estimate_jobtimer[i]

df_forecast['Dækningsbidrag (t)'] = df_forecast['Omsætning (t)']-df_forecast['Vareforbrug (t)'] -df_forecast['Lønforbrug (t)']

Forecast validation performance

Dækningsbidrag

df_forecast_val_dækningsbidrag = pd.DataFrame(df_forecast['Dækningsbidrag (t)'])
df_forecast_val_dækningsbidrag = df_forecast_val_dækningsbidrag.rename(columns={'Dækningsbidrag (t)': 'Dækningsbidrag forecast'})
df_forecast_val_dækningsbidrag['Dækningsbidrag realiseret'] = df_val_copy['Omsætning (t)']-df_val_copy['Vareforbrug (t)'] -df_val_copy['Lønforbrug (t)']
df_forecast_val_dækningsbidrag = df_forecast_val_dækningsbidrag.dropna(axis = 0)
df_forecast_val_dækningsbidrag
dækningsbidrag_val_rmse = mean_squared_error(df_forecast_val_dækningsbidrag['Dækningsbidrag realiseret'],
                                             df_forecast_val_dækningsbidrag['Dækningsbidrag forecast'],squared=False)
realiseretgennemsnitvaldækningsbidrag = np.mean(df_forecast_val_dækningsbidrag['Dækningsbidrag realiseret'])
forecastgennemsnitvaldækningsbidrag = np.mean(df_forecast_val_dækningsbidrag['Dækningsbidrag forecast'])
print(f"val RMSE dækningsbidrag: {dækningsbidrag_val_rmse} \nval gennemsnit realiseret dækningsbidrag: {realiseretgennemsnitvaldækningsbidrag} \nval gennemsnit forecast dækningsbidrag: {forecastgennemsnitvaldækningsbidrag}")

val RMSE dækningsbidrag: 529569.8530962383 
val gennemsnit realiseret dækningsbidrag: 1254598.0187500003 
val gennemsnit forecast dækningsbidrag: 1749835.2960194945

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_dækningsbidrag)
ax.set(xlabel="År og måned",
       ylabel="Dækningsbidrag (mdkk)",
       title="Dækningsbidrag forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Dækningsbidrag forecast','Dækningsbidrag realiseret'])

fig.autofmt_xdate()

plt.show()

Omsætning

df_forecast_val_omsætning = pd.DataFrame(df_forecast['Omsætning (t)'])
df_forecast_val_omsætning = df_forecast_val_omsætning.rename(columns={'Omsætning (t)': 'Omsætning forecast'})
df_forecast_val_omsætning['Omsætning realiseret'] = df_val_copy['Omsætning (t)']
df_forecast_val_omsætning = df_forecast_val_omsætning.dropna(axis = 0)
df_forecast_val_omsætning
omsætning_val_rmse = mean_squared_error(df_forecast_val_omsætning['Omsætning realiseret'],
                                        df_forecast_val_omsætning['Omsætning forecast'],squared=False)
realiseretgennemsnitvalomsætning = np.mean(df_forecast_val_omsætning['Omsætning realiseret'])
forecastgennemsnitvalomsætning = np.mean(df_forecast_val_omsætning['Omsætning forecast'])
print(f"val RMSE omsætning: {omsætning_val_rmse} \nval gennemsnit realiseret omsætning: {realiseretgennemsnitvalomsætning} \nval gennemsnit forecast omsætning: {forecastgennemsnitvalomsætning}")

val RMSE omsætning: 581748.2262998134 
val gennemsnit realiseret omsætning: 4739817.2375 
val gennemsnit forecast omsætning: 5069144.685099592

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning (mdkk)",
       title="Omsætning forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Omsætning forecast','Omsætning realiseret'])

fig.autofmt_xdate()

plt.show()

Vareforbrug

df_forecast_val_vareforbrug = pd.DataFrame(df_forecast['Vareforbrug (t)'])
df_forecast_val_vareforbrug = df_forecast_val_vareforbrug.rename(columns={'Vareforbrug (t)': 'Vareforbrug forecast'})
df_forecast_val_vareforbrug['Vareforbrug realiseret'] = df_val_copy['Vareforbrug (t)']
df_forecast_val_vareforbrug = df_forecast_val_vareforbrug.dropna(axis = 0)
df_forecast_val_vareforbrug
vareforbrug_val_rmse = mean_squared_error(df_forecast_val_vareforbrug['Vareforbrug realiseret'],
                                          df_forecast_val_vareforbrug['Vareforbrug forecast'],squared=False)
realiseretgennemsnitvalvareforbrug = np.mean(df_forecast_val_vareforbrug['Vareforbrug realiseret'])
forecastgennemsnitvalvareforbrug = np.mean(df_forecast_val_vareforbrug['Vareforbrug forecast'])
print(f"val RMSE vareforbrug: {vareforbrug_val_rmse} \nval gennemsnit realiseret vareforbrug: {realiseretgennemsnitvalvareforbrug} \nval gennemsnit forecast vareforbrug: {forecastgennemsnitvalvareforbrug}")

val RMSE vareforbrug: 344589.0722096242 
val gennemsnit realiseret vareforbrug: 1372559.0675 
val gennemsnit forecast vareforbrug: 1161600.9062907647

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug (mdkk)",
       title="Vareforbrug forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Vareforbrug forecast','Vareforbrug realiseret'])

fig.autofmt_xdate()

plt.show()

Lønforbrug

df_forecast_val_lønforbrug = pd.DataFrame(df_forecast['Lønforbrug (t)'])
df_forecast_val_lønforbrug = df_forecast_val_lønforbrug.rename(columns={'Lønforbrug (t)': 'Lønforbrug forecast'})
df_forecast_val_lønforbrug['Lønforbrug realiseret'] = df_val_copy['Lønforbrug (t)']
df_forecast_val_lønforbrug = df_forecast_val_lønforbrug.dropna(axis = 0)
df_forecast_val_lønforbrug
lønforbrug_val_rmse = mean_squared_error(df_forecast_val_lønforbrug['Lønforbrug realiseret'],
                                         df_forecast_val_lønforbrug['Lønforbrug forecast'],squared=False)
realiseretgennemsnitvallønforbrug = np.mean(df_forecast_val_lønforbrug['Lønforbrug realiseret'])
forecastgennemsnitvallønforbrug = np.mean(df_forecast_val_lønforbrug['Lønforbrug forecast'])
print(f"val RMSE lønforbrug: {lønforbrug_val_rmse} \nval gennemsnit realiseret lønforbrug: {realiseretgennemsnitvallønforbrug} \nval gennemsnit forecast lønforbrug: {forecastgennemsnitvallønforbrug}")

val RMSE lønforbrug: 146627.09003546895 
val gennemsnit realiseret lønforbrug: 2112660.15125 
val gennemsnit forecast lønforbrug: 2157708.4827893325

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug (mdkk)",
       title="Lønforbrug forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Lønforbrug forecast','Lønforbrug realiseret'])

fig.autofmt_xdate()

plt.show()

Faktureringsgrad

df_forecast_val_faktureringsgrad = pd.DataFrame(df_forecast['Faktureringsgrad (t)'])
df_forecast_val_faktureringsgrad = df_forecast_val_faktureringsgrad.rename(columns={'Faktureringsgrad (t)': 'Faktureringsgrad forecast'})
df_forecast_val_faktureringsgrad['Faktureringsgrad realiseret'] = df_val_copy['Faktureringsgrad (t)']
df_forecast_val_faktureringsgrad = df_forecast_val_faktureringsgrad.dropna(axis = 0)
df_forecast_val_faktureringsgrad
faktureringsgrad_val_rmse = mean_squared_error(df_forecast_val_faktureringsgrad['Faktureringsgrad realiseret'],
                                               df_forecast_val_faktureringsgrad['Faktureringsgrad forecast'],squared=False)
realiseretgennemsnitvalfaktureringsgrad = np.mean(df_forecast_val_faktureringsgrad['Faktureringsgrad realiseret'])
forecastgennemsnitvalfaktureringsgrad = np.mean(df_forecast_val_faktureringsgrad['Faktureringsgrad forecast'])
print(f"val RMSE faktureringsgrad: {faktureringsgrad_val_rmse} \nval gennemsnit realiseret faktureringsgrad: {realiseretgennemsnitvalfaktureringsgrad} \nval gennemsnit forecast faktureringsgrad: {forecastgennemsnitvalfaktureringsgrad}")

val RMSE faktureringsgrad: 0.027400576194875695 
val gennemsnit realiseret faktureringsgrad: 0.9190631896448261 
val gennemsnit forecast faktureringsgrad: 0.9010229005045723

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_faktureringsgrad)
ax.set(xlabel="År og måned",
       ylabel="Faktureringsgrad",
       title="Faktureringsgrad forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Faktureringsgrad forecast','Faktureringsgrad realiseret'])

fig.autofmt_xdate()

plt.show()

Jobtimer

df_forecast_val_jobtimer = pd.DataFrame(df_forecast['Jobtimer (t)'])
df_forecast_val_jobtimer = df_forecast_val_jobtimer.rename(columns={'Jobtimer (t)': 'Jobtimer forecast'})
df_forecast_val_jobtimer['Jobtimer realiseret'] = df_val_copy['Jobtimer (t)']
df_forecast_val_jobtimer = df_forecast_val_jobtimer.dropna(axis = 0)
df_forecast_val_jobtimer
jobtimer_val_rmse = mean_squared_error(df_forecast_val_jobtimer['Jobtimer realiseret'],
                                       df_forecast_val_jobtimer['Jobtimer forecast'],squared=False)
realiseretgennemsnitvaljobtimer = np.mean(df_forecast_val_jobtimer['Jobtimer realiseret'])
forecastgennemsnitvaljobtimer = np.mean(df_forecast_val_jobtimer['Jobtimer forecast'])
print(f"val RMSE jobtimer: {jobtimer_val_rmse} \nval gennemsnit realiseret jobtimer: {realiseretgennemsnitvaljobtimer} \nval gennemsnit forecast jobtimer: {forecastgennemsnitvaljobtimer}")

val RMSE jobtimer: 1271.3311383839011 
val gennemsnit realiseret jobtimer: 7557.89625 
val gennemsnit forecast jobtimer: 8710.203312605734

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Jobtimer forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Jobtimer forecast','Jobtimer realiseret'])

fig.autofmt_xdate()

plt.show()

Fakturerbar tid

df_forecast_val_fakturerbartid = pd.DataFrame(df_forecast['Fakturerbar tid (t)'])
df_forecast_val_fakturerbartid = df_forecast_val_fakturerbartid.rename(columns={'Fakturerbar tid (t)': 'Fakturerbar tid forecast'})
df_forecast_val_fakturerbartid['Fakturerbar tid realiseret'] = df_val_copy['Fakturerbar tid (t)']
df_forecast_val_fakturerbartid = df_forecast_val_fakturerbartid.dropna(axis = 0)
df_forecast_val_fakturerbartid
fakturerbartid_val_rmse = mean_squared_error(df_forecast_val_fakturerbartid['Fakturerbar tid realiseret'],
                                             df_forecast_val_fakturerbartid['Fakturerbar tid forecast'],squared=False)
realiseretgennemsnitvalfakturerbartid = np.mean(df_forecast_val_fakturerbartid['Fakturerbar tid realiseret'])
forecastgennemsnitvalfakturerbartid = np.mean(df_forecast_val_fakturerbartid['Fakturerbar tid forecast'])
print(f"val RMSE fakturerbar tid: {fakturerbartid_val_rmse} \nval gennemsnit realiseret fakturerbar tid: {realiseretgennemsnitvalfakturerbartid} \nval gennemsnit forecast fakturerbar tid: {forecastgennemsnitvalfakturerbartid}")

val RMSE fakturerbar tid: 992.1655614616113 
val gennemsnit realiseret fakturerbar tid: 6948.115 
val gennemsnit forecast fakturerbar tid: 7848.092652708551

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Fakturerbar tid forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Fakturerbar tid forecast','Fakturerbar tid realiseret'])

fig.autofmt_xdate()

plt.show()

Antal produktionsmedarbejdere

df_forecast_val_antalp = pd.DataFrame(df_forecast['Antal produktionsmedarbejdere (t)'])
df_forecast_val_antalp = df_forecast_val_antalp.rename(columns={'Antal produktionsmedarbejdere (t)': 'Antal produktionsmedarbejdere forecast'})
df_forecast_val_antalp['Antal produktionsmedarbejdere realiseret'] = df_val_copy['Antal produktionsmedarbejdere (t)']
df_forecast_val_antalp = df_forecast_val_antalp.dropna(axis = 0)
df_forecast_val_antalp
antalp_val_rmse = mean_squared_error(df_forecast_val_antalp['Antal produktionsmedarbejdere realiseret'],
                                     df_forecast_val_antalp['Antal produktionsmedarbejdere forecast'],squared=False)
realiseretgennemsnitvalantalp = np.mean(df_forecast_val_antalp['Antal produktionsmedarbejdere realiseret'])
forecastgennemsnitvalantalp = np.mean(df_forecast_val_antalp['Antal produktionsmedarbejdere forecast'])
print(f"val RMSE antal produktionsmedarbejdere: {antalp_val_rmse} \nval gennemsnit realiseret antal produktionsmedarbejdere: {realiseretgennemsnitvalantalp} \nval gennemsnit forecast antal produktionsmedarbejdere: {forecastgennemsnitvalantalp}")

val RMSE antal produktionsmedarbejdere: 1.2886441318434514 
val gennemsnit realiseret antal produktionsmedarbejdere: 65.625 
val gennemsnit forecast antal produktionsmedarbejdere: 65.73437323242807

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_val_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Antal produktionsmedarbejdere forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

ax.legend(['Antal produktionsmedarbejdere forecast','Antal produktionsmedarbejdere realiseret'])

fig.autofmt_xdate()

plt.show()

Forecast test performance

Dækningsbidrag

df_forecast_test_dækningsbidrag = pd.DataFrame(df_forecast['Dækningsbidrag (t)'])
df_forecast_test_dækningsbidrag = df_forecast_test_dækningsbidrag.rename(columns={'Dækningsbidrag (t)': 'Dækningsbidrag forecast'})
df_forecast_test_dækningsbidrag['Dækningsbidrag realiseret'] = df_test_copy['Omsætning (t)']-df_test_copy['Vareforbrug (t)'] -df_test_copy['Lønforbrug (t)']
df_forecast_test_dækningsbidrag = df_forecast_test_dækningsbidrag.dropna(axis = 0)
df_forecast_test_dækningsbidrag
dækningsbidrag_test_rmse = mean_squared_error(df_forecast_test_dækningsbidrag['Dækningsbidrag realiseret'],
                                             df_forecast_test_dækningsbidrag['Dækningsbidrag forecast'],squared=False)
realiseretgennemsnittestdækningsbidrag = np.mean(df_forecast_test_dækningsbidrag['Dækningsbidrag realiseret'])
forecastgennemsnittestdækningsbidrag = np.mean(df_forecast_test_dækningsbidrag['Dækningsbidrag forecast'])
print(f"Test RMSE dækningsbidrag: {dækningsbidrag_test_rmse} \nTest gennemsnit realiseret dækningsbidrag: {realiseretgennemsnittestdækningsbidrag} \nTest gennemsnit forecast dækningsbidrag: {forecastgennemsnittestdækningsbidrag}")

Test RMSE dækningsbidrag: 243255.92491748283 
Test gennemsnit realiseret dækningsbidrag: 1490875.2999999996 
Test gennemsnit forecast dækningsbidrag: 1733309.1910883554

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_dækningsbidrag)
ax.set(xlabel="År og måned",
       ylabel="Dækningsbidrag (mdkk)",
       title="Dækningsbidrag forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Dækningsbidrag forecast','Dækningsbidrag realiseret'])

fig.autofmt_xdate()

plt.show()

Omsætning

df_forecast_test_omsætning = pd.DataFrame(df_forecast['Omsætning (t)'])
df_forecast_test_omsætning = df_forecast_test_omsætning.rename(columns={'Omsætning (t)': 'Omsætning forecast'})
df_forecast_test_omsætning['Omsætning realiseret'] = df_test_copy['Omsætning (t)']
df_forecast_test_omsætning = df_forecast_test_omsætning.dropna(axis = 0)
df_forecast_test_omsætning
omsætning_test_rmse = mean_squared_error(df_forecast_test_omsætning['Omsætning realiseret'],
                                        df_forecast_test_omsætning['Omsætning forecast'],squared=False)
realiseretgennemsnittestomsætning = np.mean(df_forecast_test_omsætning['Omsætning realiseret'])
forecastgennemsnittestomsætning = np.mean(df_forecast_test_omsætning['Omsætning forecast'])
print(f"Test RMSE omsætning: {omsætning_test_rmse} \nTest gennemsnit realiseret omsætning: {realiseretgennemsnittestomsætning} \nTest gennemsnit forecast omsætning: {forecastgennemsnittestomsætning}")

Test RMSE omsætning: 315538.0667503667 
Test gennemsnit realiseret omsætning: 5606485.366666666 
Test gennemsnit forecast omsætning: 5596969.98922411

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_omsætning)
ax.set(xlabel="År og måned",
       ylabel="Omsætning (mdkk)",
       title="Omsætning forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Omsætning forecast','Omsætning realiseret'])

fig.autofmt_xdate()

plt.show()

Vareforbrug

df_forecast_test_vareforbrug = pd.DataFrame(df_forecast['Vareforbrug (t)'])
df_forecast_test_vareforbrug = df_forecast_test_vareforbrug.rename(columns={'Vareforbrug (t)': 'Vareforbrug forecast'})
df_forecast_test_vareforbrug['Vareforbrug realiseret'] = df_test_copy['Vareforbrug (t)']
df_forecast_test_vareforbrug = df_forecast_test_vareforbrug.dropna(axis = 0)
df_forecast_test_vareforbrug
vareforbrug_test_rmse = mean_squared_error(df_forecast_test_vareforbrug['Vareforbrug realiseret'],
                                          df_forecast_test_vareforbrug['Vareforbrug forecast'],squared=False)
realiseretgennemsnittestvareforbrug = np.mean(df_forecast_test_vareforbrug['Vareforbrug realiseret'])
forecastgennemsnittestvareforbrug = np.mean(df_forecast_test_vareforbrug['Vareforbrug forecast'])
print(f"Test RMSE vareforbrug: {vareforbrug_test_rmse} \nTest gennemsnit realiseret vareforbrug: {realiseretgennemsnittestvareforbrug} \nTest gennemsnit forecast vareforbrug: {forecastgennemsnittestvareforbrug}")

Test RMSE vareforbrug: 429772.0838960463 
Test gennemsnit realiseret vareforbrug: 1627220.11 
Test gennemsnit forecast vareforbrug: 1257374.7861916393

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_vareforbrug)
ax.set(xlabel="År og måned",
       ylabel="Vareforbrug (mdkk)",
       title="Vareforbrug forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Vareforbrug forecast','Vareforbrug realiseret'])

fig.autofmt_xdate()

plt.show()

Lønforbrug

df_forecast_test_lønforbrug = pd.DataFrame(df_forecast['Lønforbrug (t)'])
df_forecast_test_lønforbrug = df_forecast_test_lønforbrug.rename(columns={'Lønforbrug (t)': 'Lønforbrug forecast'})
df_forecast_test_lønforbrug['Lønforbrug realiseret'] = df_test_copy['Lønforbrug (t)']
df_forecast_test_lønforbrug = df_forecast_test_lønforbrug.dropna(axis = 0)
df_forecast_test_lønforbrug
lønforbrug_test_rmse = mean_squared_error(df_forecast_test_lønforbrug['Lønforbrug realiseret'],
                                         df_forecast_test_lønforbrug['Lønforbrug forecast'],squared=False)
realiseretgennemsnittestlønforbrug = np.mean(df_forecast_test_lønforbrug['Lønforbrug realiseret'])
forecastgennemsnittestlønforbrug = np.mean(df_forecast_test_lønforbrug['Lønforbrug forecast'])
print(f"Test RMSE lønforbrug: {lønforbrug_test_rmse} \nTest gennemsnit realiseret lønforbrug: {realiseretgennemsnittestlønforbrug} \nTest gennemsnit forecast lønforbrug: {forecastgennemsnittestlønforbrug}")

Test RMSE lønforbrug: 211669.99986430234 
Test gennemsnit realiseret lønforbrug: 2488389.9566666665 
Test gennemsnit forecast lønforbrug: 2606286.0119441156

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_lønforbrug)
ax.set(xlabel="År og måned",
       ylabel="Lønforbrug (mdkk)",
       title="Lønforbrug forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Lønforbrug forecast','Lønforbrug realiseret'])

fig.autofmt_xdate()

plt.show()

Faktureringsgrad

df_forecast_test_faktureringsgrad = pd.DataFrame(df_forecast['Faktureringsgrad (t)'])
df_forecast_test_faktureringsgrad = df_forecast_test_faktureringsgrad.rename(columns={'Faktureringsgrad (t)': 'Faktureringsgrad forecast'})
df_forecast_test_faktureringsgrad['Faktureringsgrad realiseret'] = df_test_copy['Faktureringsgrad (t)']
df_forecast_test_faktureringsgrad = df_forecast_test_faktureringsgrad.dropna(axis = 0)
df_forecast_test_faktureringsgrad
faktureringsgrad_test_rmse = mean_squared_error(df_forecast_test_faktureringsgrad['Faktureringsgrad realiseret'],
                                               df_forecast_test_faktureringsgrad['Faktureringsgrad forecast'],squared=False)
realiseretgennemsnittestfaktureringsgrad = np.mean(df_forecast_test_faktureringsgrad['Faktureringsgrad realiseret'])
forecastgennemsnittestfaktureringsgrad = np.mean(df_forecast_test_faktureringsgrad['Faktureringsgrad forecast'])
print(f"Test RMSE faktureringsgrad: {faktureringsgrad_test_rmse} \nTest gennemsnit realiseret faktureringsgrad: {realiseretgennemsnittestfaktureringsgrad} \nTest gennemsnit forecast faktureringsgrad: {forecastgennemsnittestfaktureringsgrad}")

Test RMSE faktureringsgrad: 0.008614368635692074 
Test gennemsnit realiseret faktureringsgrad: 0.8956489793163481 
Test gennemsnit forecast faktureringsgrad: 0.9010229005045723

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_faktureringsgrad)
ax.set(xlabel="År og måned",
       ylabel="Faktureringsgrad",
       title="Faktureringsgrad forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Faktureringsgrad forecast','Faktureringsgrad realiseret'])

fig.autofmt_xdate()

plt.show()

Jobtimer

df_forecast_test_jobtimer = pd.DataFrame(df_forecast['Jobtimer (t)'])
df_forecast_test_jobtimer = df_forecast_test_jobtimer.rename(columns={'Jobtimer (t)': 'Jobtimer forecast'})
df_forecast_test_jobtimer['Jobtimer realiseret'] = df_test_copy['Jobtimer (t)']
df_forecast_test_jobtimer = df_forecast_test_jobtimer.dropna(axis = 0)
df_forecast_test_jobtimer
jobtimer_test_rmse = mean_squared_error(df_forecast_test_jobtimer['Jobtimer realiseret'],
                                       df_forecast_test_jobtimer['Jobtimer forecast'],squared=False)
realiseretgennemsnittestjobtimer = np.mean(df_forecast_test_jobtimer['Jobtimer realiseret'])
forecastgennemsnittestjobtimer = np.mean(df_forecast_test_jobtimer['Jobtimer forecast'])
print(f"Test RMSE jobtimer: {jobtimer_test_rmse} \nTest gennemsnit realiseret jobtimer: {realiseretgennemsnittestjobtimer} \nTest gennemsnit forecast jobtimer: {forecastgennemsnittestjobtimer}")

Test RMSE jobtimer: 1491.3785516025166 
Test gennemsnit realiseret jobtimer: 8997.833333333334 
Test gennemsnit forecast jobtimer: 10131.262319523279

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_jobtimer)
ax.set(xlabel="År og måned",
       ylabel="Jobtimer",
       title="Jobtimer forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Jobtimer forecast','Jobtimer realiseret'])

fig.autofmt_xdate()

plt.show()

Fakturerbar tid

df_forecast_test_fakturerbartid = pd.DataFrame(df_forecast['Fakturerbar tid (t)'])
df_forecast_test_fakturerbartid = df_forecast_test_fakturerbartid.rename(columns={'Fakturerbar tid (t)': 'Fakturerbar tid forecast'})
df_forecast_test_fakturerbartid['Fakturerbar tid realiseret'] = df_test_copy['Fakturerbar tid (t)']
df_forecast_test_fakturerbartid = df_forecast_test_fakturerbartid.dropna(axis = 0)
df_forecast_test_fakturerbartid
fakturerbartid_test_rmse = mean_squared_error(df_forecast_test_fakturerbartid['Fakturerbar tid realiseret'],
                                             df_forecast_test_fakturerbartid['Fakturerbar tid forecast'],squared=False)
realiseretgennemsnittestfakturerbartid = np.mean(df_forecast_test_fakturerbartid['Fakturerbar tid realiseret'])
forecastgennemsnittestfakturerbartid = np.mean(df_forecast_test_fakturerbartid['Fakturerbar tid forecast'])
print(f"Test RMSE fakturerbar tid: {fakturerbartid_test_rmse} \nTest gennemsnit realiseret fakturerbar tid: {realiseretgennemsnittestfakturerbartid} \nTest gennemsnit forecast fakturerbar tid: {forecastgennemsnittestfakturerbartid}")

Test RMSE fakturerbar tid: 1371.3025463947224 
Test gennemsnit realiseret fakturerbar tid: 8061.5 
Test gennemsnit forecast fakturerbar tid: 9128.499360909545

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_fakturerbartid)
ax.set(xlabel="År og måned",
       ylabel="Fakturerbar tid",
       title="Fakturerbar tid forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)
fmt_month = mdates.MonthLocator()
ax.xaxis.set_major_locator(fmt_month)

ax.legend(['Fakturerbar tid forecast','Fakturerbar tid realiseret'])

fig.autofmt_xdate()

plt.show()

Antal produktionsmedarbejdere

df_forecast_test_antalp = pd.DataFrame(df_forecast['Antal produktionsmedarbejdere (t)'])
df_forecast_test_antalp = df_forecast_test_antalp.rename(columns={'Antal produktionsmedarbejdere (t)': 'Antal produktionsmedarbejdere forecast'})
df_forecast_test_antalp['Antal produktionsmedarbejdere realiseret'] = df_test_copy['Antal produktionsmedarbejdere (t)']
df_forecast_test_antalp = df_forecast_test_antalp.dropna(axis = 0)
df_forecast_test_antalp
antalp_test_rmse = mean_squared_error(df_forecast_test_antalp['Antal produktionsmedarbejdere realiseret'],
                                     df_forecast_test_antalp['Antal produktionsmedarbejdere forecast'],squared=False)
realiseretgennemsnittestantalp = np.mean(df_forecast_test_antalp['Antal produktionsmedarbejdere realiseret'])
forecastgennemsnittestantalp = np.mean(df_forecast_test_antalp['Antal produktionsmedarbejdere forecast'])
print(f"Test RMSE antal produktionsmedarbejdere: {antalp_test_rmse} \nTest gennemsnit realiseret antal produktionsmedarbejdere: {realiseretgennemsnittestantalp} \nTest gennemsnit forecast antal produktionsmedarbejdere: {forecastgennemsnittestantalp}")

Test RMSE antal produktionsmedarbejdere: 0.6454623574488373 
Test gennemsnit realiseret antal produktionsmedarbejdere: 72.33333333333333 
Test gennemsnit forecast antal produktionsmedarbejdere: 72.92981239183634

fig, ax = plt.subplots(figsize=(10, 6))
ax.plot(df_forecast_test_antalp)
ax.set(xlabel="År og måned",
       ylabel="Antal produktionsmedarbejdere",
       title="Antal produktionsmedarbejdere forecast vs realiseret")

ax.format_xdata = mdates.DateFormatter('%Y-%m')
ax.grid(True)

 


ax.legend(['Antal produktionsmedarbejdere forecast','Antal produktionsmedarbejdere realiseret'])

fig.autofmt_xdate()

plt.show()